Software Algorithms to Coordinate and Improve Voltage Sag ...

Clemson UniversityTigerPrints

All Dissertations Dissertations

5-2010

Software Algorithms to Coordinate and ImproveVoltage Sag Ridethrough Capabilities ofNetworked Industrial ProcessesOwen ParksClemson University, [email protected]

Follow this and additional works at: https://tigerprints.clemson.edu/all_dissertations

Part of the Electrical and Computer Engineering Commons

This Dissertation is brought to you for free and open access by the Dissertations at TigerPrints. It has been accepted for inclusion in All Dissertations byan authorized administrator of TigerPrints. For more information, please contact [email protected].

Recommended CitationParks, Owen, "Software Algorithms to Coordinate and Improve Voltage Sag Ridethrough Capabilities of Networked IndustrialProcesses" (2010). All Dissertations. 509.https://tigerprints.clemson.edu/all_dissertations/509

https://tigerprints.clemson.edu?utm_source=tigerprints.clemson.edu%2Fall_dissertations%2F509&utm_medium=PDF&utm_campaign=PDFCoverPages

https://tigerprints.clemson.edu/all_dissertations?utm_source=tigerprints.clemson.edu%2Fall_dissertations%2F509&utm_medium=PDF&utm_campaign=PDFCoverPages

https://tigerprints.clemson.edu/dissertations?utm_source=tigerprints.clemson.edu%2Fall_dissertations%2F509&utm_medium=PDF&utm_campaign=PDFCoverPages

https://tigerprints.clemson.edu/all_dissertations?utm_source=tigerprints.clemson.edu%2Fall_dissertations%2F509&utm_medium=PDF&utm_campaign=PDFCoverPages

http://network.bepress.com/hgg/discipline/266?utm_source=tigerprints.clemson.edu%2Fall_dissertations%2F509&utm_medium=PDF&utm_campaign=PDFCoverPages

https://tigerprints.clemson.edu/all_dissertations/509?utm_source=tigerprints.clemson.edu%2Fall_dissertations%2F509&utm_medium=PDF&utm_campaign=PDFCoverPages

mailto:[email protected]

SOFTWARE ALGORITHMS TO COORDINATE AND IMPROVEVOLTAGE SAG RIDETHROUGH CAPABILITIES OF

NETWORKED INDUSTRIAL PROCESSES

A DissertationPresented to

the Graduate School ofClemson University

In Partial Fulfillmentof the Requirements for the Degree

Doctor of PhilosophyElectrical Engineering

byOwen C. ParksMay 2010

Accepted by:Dr. Michael A. Bridgwood, Committee Chair

Dr. E. Randolph Collins, Jr.Dr. Adly A. Girgis

Dr. James K. Peterson

ABSTRACT

For those who design, operate, and troubleshoot industrial processes, electric

power quality is a subject that requires much consideration. Processes that use elec-

tronic sensors, actuators, and computation devices are heavily reliant on a stable,

consistent input power source. When a power quality event such as a voltage fluctu-

ation occurs, automation equipment often behaves unpredictably and causes process

malfunction or failure.

Because industrial power consumers often blame their electric utility for these

events, some utilities offer process susceptibility studies as a service for their cus-

tomers. During a typical study, utility technicians and engineers perform in-house

tests on suspect components or systems using voltage sag generating equipment.

These tests determine device malfunction thresholds and establish an event failure

timeline. Test results provide data for applying mitigation solutions, where the most

critical or susceptible loads receive a higher priority for improvement. While effective,

this approach often requires the addition of costly hardware.

This study presents novel software algorithms that coordinate and improve process

ridethrough capabilities of network connected industrial processes. An add-on PC in-

terfacing with an automation network executes a routine that detects voltage sags,

performs a fast measurement of sag parameters, and determines an expected process

response. Rather than implement a ‘cure all’ reaction for every disturbance scenario,

mitigation routines are executed based upon the expected response. Underlying de-

sign constraints of this study are to minimize or avoid the installation of conventional

ridethrough hardware and adhere to a software architecture that is unintrusive to

existing controllers.

Voltage sag detection is performed with a real-time analysis of incoming voltages

and is triggered from RMS voltage derivative threshold crossings. Having recognized

the presence of a voltage sag, the algorithm determines the sag magnitude with a

ii

peak detection method, and can associate the measured magnitude/phase combina-

tion with previously recorded process data. Either the sag characteristics or historical

process response data is then analyzed to determine the expected process response.

Sags that can potentially force motor drives to trip offline cause the process to re-

spond to an expected shutdown. Voltage sag magnitude/phasing combinations that

have been shown to cause no process disruption are ignored. Combinations which

have caused only instrument signal corruption and significant process variable devi-

ations trigger the mitigation routine to address faulted control signals only. Drive

fault mitigation responses consist of a software-only drive coast routine and an im-

proved drive coast routine requiring the addition of basic switching hardware. Out

of tolerance process errors are mitigated with output control command substitution

or input signal substitution routines.

Verification of software functionality is achieved with an experimental automated

process - - a textile unwind/rewind system that operates at a controlled linespeed

and tension. Detailed analysis and simulation is performed on both component and

system-wide levels. Unmitigated and mitigated process voltage sag responses are

recorded and matched with the theoretical process model. Although customization is

required to apply the algorithms to the specific design of the textile tension control

process, experimentation with this test bed system serves as a satisfactory proof of

concept for the software routines. As a result, the methods developed in this study

can improve the task of process power quality mitigation by customizing solutions for

individual processes, avoiding the application of power quality mitigation solutions

where they are not required, coordinating corrective actions by utilizing existing au-

tomation network functionality, and ultimately reducing the need for costly hardware

installation and maintenance.

iii

DEDICATION

Dedicated to the memory of Katherine ‘Kay’ Parks, 1916-2006.

iv

ACKNOWLEDGEMENTS

Although doctoral research requires intense individual study and self-motivated

efforts, this work would not be a reality without the assistance of others. I am deeply

indebted to:

my parents, Robert and Virginia Parks, for their steadfast love and nurturing

throughout my entire life;

Dr. Michael A. Bridgwood, my research advisor, for his guidance and wisdom

during all of my academic pursuits at Clemson;

Dr. E. Randolph Collins, Jr., Dr. Adly A. Girgis, and Dr. James K. Peterson for

their advice and assistance while serving as research committee members;

my colleagues, both past and present, in the Clemson University Power Quality

Laboratory for their aid and companionship during the countless hours spent on

coursework and research;

the people of Schneider Electric/Square D Company of Seneca, SC for their un-

wavering encouragement during the final stages of my graduate work; and

Robert L. Morgan and Duke Power Company for their generous support, and for

providing essential insight into the practical applications of this research.

Owen C. Parks

May 2010

v

TABLE OF CONTENTS

Page

TITLE PAGE . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . i

ABSTRACT . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . ii

DEDICATION . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . iv

ACKNOWLEDGEMENTS. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . v

LIST OF TABLES . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . viii

LIST OF FIGURES. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . x

CHAPTER

1. INTRODUCTION AND INVESTIGATION SUMMARY. . . . . . . . . . . . . . . . 1

Power Quality Background . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1Research Summary . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 11

2. OPERATION AND RESPONSE OF EXPERIMENTAL PROCESS . . . . 14

Topology and General Overview . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 14Simulation of Test Stand Behavior . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 21Process Response to Voltage Sags . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 26Summary. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 50

3. SOFTWARE DESIGN . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 53

Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 53Communications and Interfacing . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 53Existing System Modifications . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 56Voltage Sag Detection . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 60Data Manipulation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 65Event Analysis and Association . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 67Mitigation Routines . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 75Summary. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 85

vi

Table of Contents (Continued)

Page

4. EXPERIMENTAL RESULTS . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 87

Execution Times . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 87Performance of Sag Detection, Sag Measurement, andEvent Association . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 91Mitigation Algorithms . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 98Summary. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 128

5. CONCLUSIONS AND FUTURE WORK. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 130

Investigation Conclusions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 130Future Research Opportunities . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 133

APPENDICES . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 138

A. Textile Tension Controller Equipment List . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 139B. Textile Tension Controller Mechanical Analysis . . . . . . . . . . . . . . . . . . . . . . . . . . 140C. Determination of Physical Constants . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 148D. Matlab Process Response Simulation Code . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 157E. Instrumentation Voltage Sag Response Data . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 180F. AC Motor Drive Voltage Sag Response Data . . . . . . . . . . . . . . . . . . . . . . . . . . . . 182G. LabView Code For Data Read and Profibus Functions . . . . . . . . . . . . . . . . . . 185H. PLC Ladder Code . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 189I. Additional Ridethrough PC Software Flowcharts . . . . . . . . . . . . . . . . . . . . . . . . 208J. Main LabView Code For Ridethrough PC . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 214

BIBLIOGRAPHY . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 228

vii

LIST OF TABLES

Table Page

2.1. Event schedule for implementation of improved instrument voltagesag response model in simulation. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 36

2.2. Tension stand hardware that showed no dropout response. . . . . . . . . . . . . . 43

3.1. Ridethrough PC control word structure. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 58

3.2. Ridethrough PC status word structure. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 58

3.3. Data transferred from PLC to ridethrough PC. . . . . . . . . . . . . . . . . . . . . . . . . 66

3.4. Data transferred from ridethrough PC to PLC. . . . . . . . . . . . . . . . . . . . . . . . . 66

3.5. Process data storage array elements. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 67

4.1. Individual operation execution times. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 88

4.2. Mitigation routine predicted execution times. . . . . . . . . . . . . . . . . . . . . . . . . . . 90

4.3. Mitigation routine experimental execution times. . . . . . . . . . . . . . . . . . . . . . . 90

4.4. Actual and theoretical voltage sag detection times.. . . . . . . . . . . . . . . . . . . . . 92

A.1. Tension control system hardware information. . . . . . . . . . . . . . . . . . . . . . . . . . . 139

A.2. Programming and monitoring software. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 139

B.1. Tension control system mechanical constants. . . . . . . . . . . . . . . . . . . . . . . . . . . 141

C.1. Measured roller radii. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 148

C.2. Bare roller inertias. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 150

C.3. Rolling friction constants. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 152

C.4. Textile constants. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 152

C.5. First order device constants. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 154

C.6. Induction motor equivalent circuit parameters. . . . . . . . . . . . . . . . . . . . . . . . . . 155

viii

List of Tables (Continued)

Table Page

C.7. Volts/Hertz drive constants. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 156

E.1. Ridethrough times obtained from sag testing of tension sensor. . . . . . . . . 180

E.2. Ridethrough times obtained from sag testing of payoff tachometer. . . . . 180

E.3. Ridethrough times obtained from sag testing of takeup tachometer. . . . 180

E.4. Maximum recovery times obtained from instrument sag tests.. . . . . . . . . . 181

F.1. AC motor drive sag response data. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 183

F.2. AC motor drive sag response data (continued).. . . . . . . . . . . . . . . . . . . . . . . . . 184

ix

LIST OF FIGURES

Figure Page

1.1. Example voltage sag waveform (20%, 300ms). . . . . . . . . . . . . . . . . . . . . . . . . . . 3

1.2. ITI (CBEMA) curve of typical information technology equipmenttolerances to voltage disturbances [32]. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 6

2.1. Experimental textile tension control system mechanical layout andinstrumentation scheme. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 15

2.2. Equipment configuration of experimental textile tension controlsystem. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 17

2.3. Tension control system block diagram. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 18

2.4. Controller for textile tension control system. . . . . . . . . . . . . . . . . . . . . . . . . . . . 22

2.5. Flowchart of tension controller voltage sag simulation program. . . . . . . . . 24

2.6. Textile tension control stand instrumentation voltage sag responses. . . . 28

2.7. Instrument input-output topologies for a) DC sensor fed throughseparate power supply and b) AC supplied sensor. . . . . . . . . . . . . . . . . . . . . . 28

2.8. Instrument voltage sag responses divided into segments of operation. . . . 29

2.9. Effects of variations in sag magnitude and physical input level onridethrough segment of instrument voltage sag response. . . . . . . . . . . . . . . 30

2.10. Effects of variations in sag magnitude and physical input level ondeviation segment of instrument voltage sag response. . . . . . . . . . . . . . . . . . 32

2.11. Effects of variations in sag duration on deviation segment ofinstrument voltage sag response. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 33

2.12. Effects of variations in sag magnitude and physical input level onrecovery segment of instrument voltage sag response. . . . . . . . . . . . . . . . . . . 34

2.13. Application of mathematical recovery response model to instrumentusing constant physical input level. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 35

2.14. Flowchart for simulation of instrument voltage sag response. . . . . . . . . . . . 37

x

List of Figures (Continued)

Figure Page

2.15. DC bus levels during three phase sags of magnitude 60% (left) and80% (right). . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 39

2.16. Sag duration effects on the motor drive DC bus levels. . . . . . . . . . . . . . . . . . 41

2.17. Loading effects on the motor drive DC bus decay rates. . . . . . . . . . . . . . . . . 42

2.18. Region of voltage sags for any process disturbance in textile tensioncontrol system. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 44

2.19. Region of voltage sags for AC motor drive dropout in textile tensioncontrol system. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 45

2.20. Two phase graph of AC drive dropout points and dropout boundaryestimation line. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 47

2.21. Actual and simulated textile tension controller responses duringmotor drive dropout. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 48

2.22. Region of voltage sags for instrumentation malfunction in textiletension control system. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 49

2.23. Comparison of actual and simulated textile tension controllerresponse to 450ms, 0% sag (interruption) affecting theinstrumentation only. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 51

3.1. Ridethrough PC interface structure. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 54

3.2. Screenshot of Ridethrough PC user interface. . . . . . . . . . . . . . . . . . . . . . . . . . . 57

3.3. Flowchart of added PLC subroutine. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 59

3.4. Flowchart of voltage sag measurement and detection routine. . . . . . . . . . . 61

3.5. Simulated voltage sag (20%, 90) with RMS voltage, RMS voltagederivative, and threshold signals. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 63



xi


Figure Page



3.10. Flowchart of event analysis in ridethrough PC. . . . . . . . . . . . . . . . . . . . . . . . . . 69

3.11. Theoretical response of peak detect routine to 40%, 0 point onwave voltage sag. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 71

3.12. Theoretical response of peak detect routine to 40%, 90 point onwave voltage sag. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 71

3.13. Flowchart for identification of expected process voltage sag responseusing process history.. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 73

3.14. Flowchart for identification of expected process voltage sag responseusing threshold calculations. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 74

3.15. Flowchart of software momentary drive coast mitigation routine. . . . . . . 76

3.16. Flowchart of process restarting sequence following coast mitigationroutine.. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 77

3.17. Theoretical process response to AC drive coast mitigation routines.. . . . 78

3.18. Flowchart of hardware assisted momentary drive coast mitigationroutine.. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 79

3.19. Add-on circuit for hardware assisted momentary drive coastmitigation routine. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 80

3.20. Flowchart of control signal substitution mitigation routine. . . . . . . . . . . . . 82

3.21. Theoretical process response to control signal substitution andselective signal substitution routines.. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 83

3.22. Flowchart of selective instrument signal substitution mitigationroutine.. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 84

4.1. Order of operations for analysis of execution times. . . . . . . . . . . . . . . . . . . . . 87

xii


Figure Page

4.2. Experimental detection response to 20%, 90 point on wave voltagesag.. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 92





4.7. Experimental magnitude measurement response to 40%, 0 point onwave voltage sag. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 96

4.8. Experimental magnitude measurement response to 40%, 90 pointon wave voltage sag. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 96

4.9. Experimental and simulated unmitigated process response resultingin drive dropout (1.0m/s, 30N process setpoints; 0%, 450ms, 3φvoltage interruption). . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 100

4.10. Experimental and simulated process response using softwaremomentary drive coast algorithm (1.0m/s, 30N process setpoints;0%, 450ms, 3φ voltage interruption). . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 101

4.11. Experimental and simulated DC bus magnitude, motor output, andcontrol signal replace command during software momentary drivecoast algorithm (1.0m/s, 30N process setpoints; 0%, 450ms, 3φvoltage interruption). . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 102

4.12. Experimental and simulated process response demonstratingunsuccessful process recovery during application of softwaremomentary drive coast algorithm (1.0m/s, 30N process setpoints;0%, 450ms, 3φ voltage interruption). . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 103

4.13. Experimental and simulated DC bus magnitude, motor output,and control signal replace commands during unsuccessful software

xiii


Figure Page

momentary drive coast algorithm (1.0m/s, 30N process setpoints;0%, 450ms, 3φ voltage interruption). . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 104

4.14. Experimental and simulated process response using hardwareassisted momentary drive coast algorithm (1.0m/s, 30N processsetpoints; 0%, 450ms, 3φ voltage interruption). . . . . . . . . . . . . . . . . . . . . . . . . 107

4.15. Experimental and simulated process response using hardwareassisted momentary drive coast algorithm (1.0m/s, 30N processsetpoints; 20%, 200ms, 3φ voltage sag). . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 108

4.16. Comparison of DC bus decay rates during momentary drive coastingalgorithm executions. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 109

4.17. Experimental and simulated process response unnecessarily usinghardware assisted momentary drive coast algorithm (1.0m/s, 30Nprocess setpoints; 20%, 200ms, 3φ voltage sag).. . . . . . . . . . . . . . . . . . . . . . . . 110

4.18. Experimental and simulated unmitigated process response resultingin faulted instrumentation (1.0m/s, 30N process setpoints; 0%,450ms, 1φ voltage interruption). . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 113

4.19. Experimental and simulated process response using control signalsubstitution algorithm (1.0m/s, 30N process setpoints; 0%, 450ms,1φ voltage interruption). . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 114

4.20. Experimental and simulated faulted instrument signals and controlreplace bits (1.0m/s, 30N process setpoints; 0%, 450ms, 1φ voltageinterruption). . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 115

4.21. Experimental and simulated process response unnecessarily usingcontrol signal substitution algorithm (1.0m/s, 30N processsetpoints; 0%, 50ms, 1φ voltage interruption). . . . . . . . . . . . . . . . . . . . . . . . . . 116

4.22. Experimental and simulated process response using selective signalsubstitution algorithm (1.0m/s, 30N process setpoints; 0%, 450ms,1φ voltage interruption). . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 118

4.23. Experimental and simulated faulted instrument signals and replacebits for selective signal substitution algorithm (1.0m/s, 30N processsetpoints; 0%, 450ms, 1φ voltage interruption). . . . . . . . . . . . . . . . . . . . . . . . . 119

xiv


Figure Page

4.24. Experimental and simulated unmitigated process response resultingin faulted instrumentation (2.5m/s, 50N process setpoints; 0%,450ms, 1φ voltage interruption). . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 120

4.25. Experimental and simulated process response using selective signalsubstitution algorithm (2.5m/s, 50N process setpoints; 0%, 450ms,1φ voltage interruption). . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 121

4.26. Experimental and simulated faulted instrument signals and replacebits for selective signal substitution algorithm (2.5m/s, 50N processsetpoints; 0%, 450ms, 1φ voltage interruption). . . . . . . . . . . . . . . . . . . . . . . . . 122

4.27. Experimental and simulated unmitigated process response resultingin faulted instrumentation (1.0m/s, 10N process setpoints; 40%,450ms, 1φ voltage sag). . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 124

4.28. Experimental and simulated process response using selective signalsubstitution algorithm (1.0m/s, 10N process setpoints; 40%,450ms, 1φ voltage sag). . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 125

4.29. Experimental and simulated faulted instrument signals and replacebits for selective signal substitution algorithm (1.0m/s, 10N processsetpoints; 40%, 450ms, 1φ voltage sag). . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 126

4.30. Comparison of process responses and signal substitution intervalsfor selective signal substitution algorithm executions. . . . . . . . . . . . . . . . . . . 127

B.1. Mechanical configuration of web tension control system. . . . . . . . . . . . . . . . 141

B.2. Mechanical constants for web tension control system. . . . . . . . . . . . . . . . . . . 141

B.3. Free body diagram of forces on reel #1 (takeup). . . . . . . . . . . . . . . . . . . . . . . 142

B.4. Free body diagram of forces on reel #2 (idler). . . . . . . . . . . . . . . . . . . . . . . . . . 143

B.5. Free body diagram of forces on reel #5 (payoff). . . . . . . . . . . . . . . . . . . . . . . . 144

G.1. Labview subroutine to close Profibus interface. . . . . . . . . . . . . . . . . . . . . . . . . . 185

G.2. Labview subroutine to read in data file for ridethrough PC (part 1of 2). . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 186

xv


Figure Page

G.3. Labview subroutine to read in data file for ridethrough PC (part 2of 2). . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 186

G.4. Labview subroutine for Profibus open interface (part 1 of 2). . . . . . . . . . . 187

G.5. Labview subroutine for Profibus open interface (part 2 of 2). . . . . . . . . . . 188

I.1. Flowchart for main program of ridethrough PC.. . . . . . . . . . . . . . . . . . . . . . . . 208

I.2. Flowchart for Profibus open interface routine. . . . . . . . . . . . . . . . . . . . . . . . . . . 209

I.3. Flowchart for Profibus close interface routine. . . . . . . . . . . . . . . . . . . . . . . . . . . 209

I.4. Flowchart for data file read routine. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 210

I.5. Flowchart for data file write routine. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 211

I.6. Flowchart for main loop of ridethrough PC software. . . . . . . . . . . . . . . . . . . . 212

I.7. Flowchart for Profibus read and write routine. . . . . . . . . . . . . . . . . . . . . . . . . . 213

J.1. Main Labview code - variable initialization before main loop. . . . . . . . . . . 214

J.2. Main LabView code - data read, array initialization, and hardwaresetup before main loop. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 215

J.3. Main LabView code - Profibus I/O and data monitoring. . . . . . . . . . . . . . . 216

J.4. Main LabView code - event data capture. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 217

J.5. Main LabView code - prefault average calculation. . . . . . . . . . . . . . . . . . . . . . 218

J.6. Main LabView code - data acquisition and voltage data handling. . . . . . 219

J.7. Main LabView code - voltage sag detection, minimum RMS voltage,and minimum RMS voltage derivative determination.. . . . . . . . . . . . . . . . . . 220

J.8. Main LabView code - event analysis instructions. . . . . . . . . . . . . . . . . . . . . . . 221

J.9. Main LabView code - peak voltage detection instructions. . . . . . . . . . . . . . 222

J.10. Main LabView code - expected process response identification usingrecorded process history. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 222

xvi


Figure Page

J.11. Main LabView code - expected process response identification usingformal region boundary definitions. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 223

J.12. Main LabView code - software coast agorithm instruction set. . . . . . . . . . 223

J.13. Main LabView code - hardware coast algorithm instruction set. . . . . . . . 224

J.14. Main LabView code - restarting instruction set. . . . . . . . . . . . . . . . . . . . . . . . . 224

J.15. Main LabView code - control signal substitution and status clearinginstruction set. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 225

J.16. Main LabView code - generation of ridethrough time index forselective signal substitution algorithm. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 225

J.17. Main LabView code - selective signal substitution algorithm replacebit set and reset instructions. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 226

J.18. Main LabView code - data file write and Profibus interface closefunction call after termination of main loop. . . . . . . . . . . . . . . . . . . . . . . . . . . . 227

xvii

CHAPTER 1

INTRODUCTION

AND INVESTIGATION SUMMARY

Power Quality Background

As electronics become increasingly prevalent in many applications, effective device

or process operation becomes more dependent on the quality of incoming electric

power. Test studies have shown that even commonplace devices such as digital clocks

[1,2] and personal computers [3,4] are susceptible to malfunctions caused by power

source magnitude and waveform disturbances. For some consumers of electricity,

power quality issues may not be more than an annoyance. However, because of

their scale and dependence on electronics, commercial and industrial consumers face

potentially large losses of time and money caused by power quality events.

Events and Effects

The effects of power quality events are widely dependent on their characteris-

tics. IEEE Standard 1159-1995 defines and categorizes power quality events into

three major categories of transients, waveform distortions, and voltage variations [5].

Transients are brief distortions in a waveform that do not occur periodically. A com-

mon example of a power system transient is a brief supply overvoltage caused by

utility capacitor switching, that in turn can cause equipment to fail or malfunction

[6]. Waveform distortions occur on a periodic basis, and are identifiable as a consistent

deviation from a true sine wave in a supply voltage. Waveform distortions include DC

offsets, waveform notching, harmonics, interharmonics, and electrical noise [5]. Their

effects, such as clock drift [1] and premature equipment failure due to overheating

[7] are more evident in the long term. Voltage variations occur in the steady state

as sustained undervoltages and overvoltages, and on a transient basis as voltage sags

and swells [5].

1

Consumers are susceptible to power quality events in numerous ways. For facili-

ties such as hospitals, uninterrupted power is of critical importance. This mandates

backup generation and smooth source transfer capabilities as part of the site power

distribution design [8]. Customers who are heavily reliant on information technology

equipment generally find that wiring and grounding errors are the principal causes

of their power quality problems, yet experience additional losses due to interruptions

and deep voltage sags [9]. Computers and process control equipment often utilize

internal protection against low operating voltages, which increases their sensitivity

to brief disturbances [10], and in turn leaves their users vulnerable to incur the costs

associated with their malfunctions. These costs can include factors such as lost raw

materials, lost production time, and even damaged equipment [11].

Voltage sags

Voltage sags receive close attention as troublesome power quality events, and

are considered among the most important power quality problems facing industrial

and commercial customers today [12]. Voltage sags are defined as a drop in voltage

magnitude from between 10% to 90% of the nominal value, for a duration of one-half

cycle to one minute [5]. Figure 1.1 shows an example of a 20%, 300ms voltage sag

waveform. Most frequently, voltage sags are caused by lightning induced single line

to ground faults [13]. They may also be caused by other factors such as high current

motor starting [14] and miscellaneous power system faults due to traffic/construction

accidents, animal contact, or tree contact [15]. Voltage sags have many quantifiable

characteristics, which can be influenced by factors such as loading conditions, power

system network topologies, and the presence of embedded generation [16]. In addition

to magnitude and duration, voltage sags possess characteristics such as phase angle

shift, point of wave of inception, and point of wave of recovery [17]. Furthermore,

dynamic loads such as motor operation during and after a voltage sag can influence

a sag’s transient magnitude [18]. Even though many factors contribute to voltage

sag properties, they are sometimes predictable when caused by certain events such as

remote distribution fault clearing [19].

2

Figure 1.1 Example voltage sag waveform (20%, 300ms).

Voltage Sags and Process Manufacturing

Because of their complexity, industrial processes are particularly vulnerable to

voltage sags. When a process has no response to a voltage sag, it is said to ‘ride

through’ the event. Otherwise, one of three scenarios will generally occur: the process

may shut down even though operating parameters remain within specification, the

process may shut down by exceeding process parameter failsafe levels, or the process

does not shut down but instead produces an out of specification product [20].

Many types of processes behave in this manner. Textile handling systems, for

example, are highly susceptible to voltage sag disturbances. A review of over 60

textile facilities performed by the Electric Power Research Institute (EPRI), Duke

Power, Carolina Power and Light, Georgia Power, and Northeast Utilities showed

that every segment of the textile industry is susceptible to power quality problems

[21]. Furthermore, an increased need to compete in global markets has fueled the

3

need to increase textile automation, which in turn increases vulnerability throughout

the industry [21]. Other industries that rely heavily on processes, such as electronics

manufacturing and paper production, are also highly vulnerable to voltage sags [22,

23]. In industries that rely on extrusion operations, where cleanup and restarting

are complicated procedures, costs per event can be on the order of $10,000, with 20-

25 events occurring per year [24]. Even non-continuous processes such as computer

numerical control (CNC) operations can fault or produce out of tolerance products

due to voltage sags [20].

Some commonly used devices are often blamed for process malfunctions. Ad-

justable speed drives can decrease their output speed during a voltage sag, and reac-

celeration times may be lengthened by current limitations intrinsic to the drive [25].

AC contactor motor starters can cause complete process shutdowns because of their

susceptibility to voltage sags, and are sometimes the weakest link in an entire process

[26]. Their ridethrough behavior is also sensitive to the point on wave of voltage sag

inception [27], which makes identifying their dropout threshold a complicated task.

Programmable logic controllers (PLC) have several aspects of sag vulnerability. A

PEAC study of PLC susceptibility showed that a PLC power supply can cause PLC

dropout (and subsequent process shutdown), and standard 120VAC discrete inputs

can be misinterpreted as false control signals [28]. False output signals can also be

issued by a PLC that is stricken by a voltage sag [29].

Evaluating a process’s susceptibility to voltage sags requires knowledge of both

the process mechanics and the equipment in use. Understanding process variables

and how they are affected by voltage disturbances is a critical part of an overall

study of system susceptibility [20]. For process equipment, it is recommended that

susceptibility studies establish a device dropout hierarchy, and compare that hierarchy

to the number of times per year sags of a particular magnitude and duration will occur

[30, 31]. This approach encourages troubleshooters to consider the likelihood that a

device will fail along with the economic factors that are involved in improving that

device’s ridethrough capabilities.

4

Studies of common information technology equipment have been performed by

the Information Technology Industry Council (ITI, formerly Computer & Business

Equipment Manufacturers Association), and have yielded information regarding typi-

cal responses to voltage disturbances of varying magnitudes and durations. These are

graphically represented in the ‘ITI Curve’, which is shown in figure 1.2. This curve

defines different regions of operation for disturbances of varying magnitude and dura-

tion. For voltage sags, where the nominal voltage decreases below 90%, tolerance of

at least one cycle for even the deepest sags is common. At higher magnitudes, equip-

ment can often ride through sustained sags. When a device does not ride through, it

is said to operate in the ‘no damage region’, which indicates dropout or malfunction,

but leads to no lasting damage [32].

Using a graphical representation similar to that seen in the ITI Curve, test results

on process equipment may be plotted to indicate their voltage sag susceptibility in

relation to other process devices. The availability of high power output, on-demand

voltage sag generators makes process-wide sag testing now possible. This test equip-

ment employs specialized field excitation of a synchronous generator, or connects

an intermediate tap switching transformer between source and load under test [33].

Using sag generators to test equipment is an effective way to determine device sen-

sitivity, and repeated testing over a range of magnitudes and durations can define a

device’s dropout thresholds as they relate to other devices or standards [34].

For complex processes such as those found in the textile industry, determining the

sensitivity of component parts is considered an effective method of establishing overall

process susceptibility. Solutions first aim at improving the voltage sag tolerance of

the weakest and most critical process components, with each solution implemented

following an analysis of its economic viability [35]. Device interaction can also yield

important information about process susceptibility. For example, sag response case

studies have been successful in predicting a process sag response by breaking down a

process into interconnected subsystems, then tracking the disturbance as it propagates

through them [36,37]. This helps determine each subsystem’s contribution to the

response without treating components as isolated entities.

5

ITI (CBEMA) Curve (Revised 2000)

0

100

200

300

400

500

0.000001 0.00001 0.0001 0.001 0.01 0.1 1 10 100

Duration in Cycles (c) and Seconds (s)

Per

cen

t o

f N

om

inal

Vo

ltag

e (R

MS

or

Pea

k E

qu

ival

ent)

20 ms 3 ms 1 ms 0.5 s 10 s Steady State

1 c 10 c 100 c 0.01 c 1 us

.001 c

120

140

110

90 80 70

40

Applicable to Single-Phase

120-Volt Equipment

Voltage Tolerance Envelope

Prohibited Region

No Damage Region

No Interruption In Function Region

Figure 1.2 ITI (CBEMA) curve of typical information technology equipment toler-ances to voltage disturbances [32].

6

Service, Education, and Research

Electric utilities face both challenges and opportunities when addressing power

quality issues. In a deregulated electric power industry, customer service is increas-

ingly important for utility business. This provides motivation for utilities to actively

address customer power quality issues [38]. Customer complaints may be handled by

power quality service programs that offer diagnostic and repair services that operate

on both the utility and customer sides of the power system. Additionally, promotion

of research and development is considered a valuable aspect of a utility power quality

program because laboratory results may be regularly transferred to utility technicians

to provide them with a stronger knowledge base to work from [39].

Even in a service environment hospitable to addressing customer complaints, there

exists a growing understanding that solving power quality problems is the shared

responsibility of utilities, electrical systems designers, installers, equipment manufac-

turers, and end users [40]. Equipment manufacturers must understand the power

quality environment so they may effectively comply with specified ridethrough re-

quests, or simply improve product ridethrough in general [41]. When constructing

and upgrading facilities, systems designers and process integrators who are aware of

power quality issues may alter purchasing decisions by weighing the added cost of

built-in equipment ridethrough against projected downtime costs [42]. When imple-

menting a mitigation solution, end users must take economics into consideration by

accounting for factors such as purchase costs, installation costs, maintenance costs,

and remaining unsolved downtime costs [43].

Conventional Mitigation Solutions

Power quality mitigation solutions may be implemented in many ways. From the

utility side, the number of customer voltage sags can be reduced by improving fault

prevention practices and modifying fault clearing methods [24]. Use of equipment

such as electronic tap changers can compensate for loss of voltage, but these devices

often have an operational delay which exceeds the ridethrough times of sensitive loads

[10]. A utility end approach to power quality mitigation does have limitations. For

7

example, mitigation of sags caused by faults at transmission voltages is considered im-

practical on the utility side because it requires protecting transmission lines that may

be extremely long or outside a utility’s property. For these types of sags, mitigation

solutions are best implemented on a customer or device specific level [44].

Customers may install specialized equipment to improve the quality of their power.

For very critical loads, devices such as static transfer switches (STS) or uninterruptible

power supplies (UPS) may be used. Static transfer switches can compensate for a

complete loss of power by switching to a secondary source using electronic switching

[45]. STS use requires a reliable backup power source that is capable of meeting

its downstream load demands. This source may be a standby power generator or

battery bank. Uninterruptible power supplies use battery banks for their backup

energy source, and may be switched online using an STS scheme, or remain online at

all times [46]. UPS protection works well for low power requirements, but becomes

economically unfeasible as the power demands and subsequent battery maintenance

costs increase [10].

Devices designed for voltage sag mitigation are not required to supply power

during a sustained interruption, and therefore only require stored energy to supply

loads briefly. A simple ridethrough technology is the motor-generator (MG) set, which

has relatively high efficiency and low initial cost [24]. The energy stored in a large

rotating mass keeps momentary voltage disturbances from significantly affecting the

mass’s rotational speed, and hence the output power produced by the generator. An

additional advantage of MG set usage is that clean power is produced on-site, which

blocks waveform distortions from being passed through from the utility. Even though

MG sets can provide significant ridethrough, their use is primarily seen in industrial

environments because of their size, maintenance requirements, and noisy operation

[10].

Series voltage controllers (SVC), also called dynamic voltage restorers (DVR),

use capacitive energy storage rather than mechanical. These devices are inserted up-

stream of a sensitive load, and inject variable additive voltage to compensate for drops

from nominal voltage [10]. An alternative DVR design allows for series compensation

8

at existing distribution transformers, which eliminates the need for purchasing and

installing insertion transformers [47]. SVC/DVR is an attractive means of mitiga-

tion to large customers with many sensitive loads, but are costly and cannot protect

against interruptions or against sags generated within a plant [10].

Other mitigation techniques use magnetic characteristics to mitigate power qual-

ity disturbances. Superconducting magnetic energy storage devices (SCMES) use

cryogenically cooled superconducting magnets to store energy, and deliver it in a

manner similar to a UPS. They take less space than their UPS equivalents, but are

expensive because of their cooling requirements [12]. Magnetic synthesizers, which

are typically used for large loads, convert electric energy into magnetic energy through

nonlinear chokes, then synthesize an output waveform using stored energy in satura-

tion transformers and capacitors [46]. Advanced static var compensators, which are

used for reactive power management, may also be used for voltage sag mitigation.

They employ existing equipment that is used for other purposes, but require com-

plicated controls and are adversely affected by phase angle jumps in a voltage sag

[48]. Constant voltage transformers (CVT), also called ferroresonant transformers,

use magnetic saturation properties to lessen the effects of voltage disturbances. They

are tuned to specific load requirements, and are best suited for steady loads [10]. They

can also help mitigate waveform disturbances such as harmonics and notching [49].

In practice, CVTs have been shown to be effective in protecting control equipment

and PLCs in a coordinated process [50].

On a device level, solutions may be implemented to reduce the occurrence of volt-

age sags or improve device ridethrough. High motor starting currents can be limited

by ‘soft start’ devices, or by programming a motor drive to slowly accelerate a load.

Ridethrough of induction motor drives may be improved with advanced pulse width

modulation techniques in the presence of a voltage sag [51]. Addition of extra ca-

pacitance in power supplies can improve ridethrough performance at low costs, but

requires internal modifications to existing equipment, and may cause premature rec-

tifier diode failure [52]. Process contactors in a may have their ridethrough improved

by making modifications to their magnetic circuits [26].

9

Software Mitigation Solutions

Software based solutions for avoiding a detrimental voltage sag process response

are also possible, and form the central issue of this dissertation. Upon detection

of a voltage sag, a process may switch to an alternate control algorithm, and then

return to normal operating conditions afterward [36]. This solution requires detailed

knowledge of process mechanics and controls. Additionally, a fast acting voltage sag

detector must be integrated into the process controls. Recent developments in sag

detection technology have shown that detectors may be economically manufactured

and implemented in a process to serve as a trigger for alternate control [53]. Newer

sag detector designs even utilize microprocessor controls for increased reaction times

and tolerance to steady state waveform disturbances [54]. Voltage sag detection using

dq input voltage analysis has also been shown to have fast response times, but has a

poor tolerance to waveform disturbances such as harmonics [55].

Control in the presence of sensor failure is addressed in [56-58], but requires re-

dundant sensors and a control algorithm designed to accommodate the redundancy.

Here methods such as weighting functions or majority decision are used to determine

an adopted value to feed back into a control algorithm. These involve identifying and

responding to faulted sensor outputs, yet do not account for input parameters such

as supply voltage characteristics. Examples of observer control for complex processes

are shown in [59] and [60]. This diminishes the possibility of signal corruption by volt-

age sags by reducing an overall sensor count, but requires computationally expensive

calculations and reliable input signals for implementation.

10

Research Summary

Motivation and Purpose

In recent years, manufacturers have increased their use of sensor/actuator bus

systems in factory processes. Before the advent of this technology, automated sys-

tems typically consisted of a PLC or computer controller connected to multiple input

and output signal lines, where each line is dedicated to carrying status or control

information between the controller and a single device. As process complexity grows,

the number of required input and output signal lines increases as well. System instal-

lation and maintenance costs are consequently increased due to the large number of

connections and lines that exist. These characteristics gave rise to the introduction

and widespread use of sensor/actuator communications bus systems, which use net-

working technology to connect multiple factory devices to a single communications

cable.

Access to a process communications bus creates an opportunity for developing

improved software-based voltage sag mitigation methods. The conventional approach

to improving process ridethrough focuses on adding voltage sag compensation equip-

ment to critical hardware, or issuing ‘cure-all’ commands to a process when a voltage

sag or its effects are detected. The presence of a data bus allows for the creation of

monitoring and control software that may easily interface with the communications

bus and issue alternate control and override commands in the presence of a voltage

sag. This software may also interface with the incoming power supply for inclusion of

voltage sag detection and measurement routines. In this environment, a customized

approach to voltage sag mitigation may be used, where measured voltage sag charac-

teristics influence the method of mitigation response. Connections with the data bus

also help facilitate recording a voltage sag response history for use in determining an

expected process response.

Voltage Sag Ridethrough Software

In this investigation, coordinated voltage sag response software is designed and

implemented in a test bed process. The software design calls for avoiding the addition

11

of conventional ridethrough hardware and minimizing modifications to existing con-

trol algorithms. It resides on a standalone PC that interfaces with both the process

communications bus and the incoming power supply. The software performs the

following functions:

1. Continuously monitors critical process signals and three phase line voltages.

2. Detects voltage sag inception and extinction using a fast detection algorithm.

3. Performs sag measurements (magnitude, duration) while the sag is in progress.

4. Records unmitigated process responses for future reference.

5. Uses voltage sag measurements to determine an expected process response by

either accessing recorded process data or calculating an expected response using

mathematical models of process behavior.

6. Determines an appropriate mitigation response based on both the expected

response and preselected user input.

7. Issues commands to the process controller (PLC) that override the existing

control algorithm and mitigate the process voltage sag response.

8. Returns the process to normal operating conditions after the voltage sag has

ended.

For the purposes of experimental verification, the software is applied to a dedi-

cated test bed process. This process is an integrated textile tension control system.

Implementation of the software in an experimental environment requires a high level

of customization, and therefore calls for analysis and simulation of the unmitigated

and mitigated process responses. Understanding individual device behavior is also

critical when implementing a process-wide mitigation strategy. This requires the de-

velopment of subsystem device behavior models for use in a simulation environment

and in the software mitigation routines themselves.

12

The details of this investigation are divided into three main sections. First, the

experimental textile tension controller design and modeling is presented. In this fo-

rum, mathematical models and simulation are discussed. The unmitigated voltage

sag process response is matched with the theoretical model obtained from simulation.

This is followed by an explanation of the design and theoretical operation of the soft-

ware routines used to mitigate the tension controller’s voltage sag responses. Lastly,

an experimental evaluation of the effectiveness of the software algorithms is presented

as they are applied to the process.

13

CHAPTER 2

OPERATION AND RESPONSE

OF EXPERIMENTAL PROCESS

This chapter describes the design, construction, and response of a model textile

tension control process that is used for applications of software based ridethrough

algorithms. A thorough mathematical analysis is included, followed by a discussion

of process simulation. An investigation into the unmitigated voltage sag response of

both individual components and the entire process ensues.

Topology and General Overview

A textile handling workstation was constructed for the purposes of determining

process responses to power quality events and experimentating with mitigation strate-

gies. The system is a multiple input-multiple output (MIMO) textile winding and

tension controller, with a design philosophy of being self-contained and dedicated to

research experiments. Textile was chosen as the process medium because of the preva-

lence of textile manufacturing plants in North and South Carolina and the associated

possibility of nearby industry implementation of methods developed in this research.

Because it does not function as a critical part of an existing manufacturing facility,

the textile tension controller may be altered and experimented with in greater detail

than a production-level process.

The mechanical layout of the textile tension controller is shown in Figure 2.1.

Two AC motors actuate the payoff and takeup reels where the textile is accumulated.

A stationary load cell measures web tension, tachometers measure reel speeds, and

discrete proximity sensors determine end of cycle package accumulation on the takeup

and payoff reels. Secondary analog instruments are included to measure process values

when a primary analog instrument is affected by a voltage sag. The secondary load

cell tension sensor and spool tachometers are not connected to the control system

and serve solely as a source of backup measurement.

14

PRIMARY TENSIONSENSOR

SECONDARYTENSIONSENSOR

IDLER ROLLER PAYOFF DRIVESPOOL

TAKEUP DRIVESPOOL

TENSIONSIGNAL

CONDITIONER

PROX. SENSOR -END OF CYCLE

PACKAGE BUILD

PROX. SENSOR -END OF CYCLE

PACKAGE BUILD

DIRECTION OF TEXTILETRAVEL

TAKEUP SPEEDSIGNAL

CONDITIONER

PAYOFF SPEEDSIGNAL

CONDITIONER

SECONDARYTENSIONSIGNAL

CONDITIONER

PRIMARY ANDSECONDARY

TACHOMETERPROBES

PRIMARY ANDSECONDARY

TACHOMETERPROBES

TEXTILE WEB

SE

CO

ND

AR

Y T

AK

EU

P S

PE

ED

SIG

NA

L

PR

IMAR

Y T

AKE

UP

SP

EED

SIG

NA

L

FULL

SP

OO

L S

IGN

AL -

TAK

EU

P

PR

IMAR

Y T

EN

SIO

N S

IGN

AL

SEC

ON

DA

RY

TEN

SIO

N S

IGN

AL

EM

PTY

SP

OO

L S

IGN

AL

- PA

YOFF

PR

IMA

RY

PAY

OFF

SP

EED

SIG

NA

L

SE

CO

ND

AR

Y P

AY

OFF

SP

EE

D S

IGN

AL

Figure 2.1 Experimental textile tension control system mechanical layout and instru-mentation scheme.

15

Figure 2.2 details the electrical interconnections of the textile tension controller.

The 0-10VDC primary instrument feedback signals are connected to an I/O block

which samples the analog signals and converts them to signed word data values. The

data is transferred along a Siemens Profibus DP network into a Siemens PLC. The

PLC receives user defined start/stop and setpoint commands from a standalone PC

running a LabView human-machine interface (HMI) program. The PLC executes

the main control algorithm and feeds the output commands back into the Profibus

network, where they are received by two Siemens AC motor drives. The drives each

control an AC induction motor, which is directly connected to the drive reels. A

Siemens AS-I bus transfers end of cycle signals generated by the spool accumulation

sensors and operates independent of the Profibus network. Programming and data

bus monitoring interfaces are also connected to the system. For a complete system

equipment list the reader is referred to Appendix A.

Textile tension control stand design is based on duplicating elements commonly

found in industrial processes. The use of load cell tension measurement was chosen

over a dancer arm system because of its mechanical simplicity. A comparison of dancer

arm and load cell control strategies is presented in [61], and states that although

differences exist in modeling and control of each system, there are no significant

advantages of employing either method. A Profibus DP data bus network was chosen

for device interconnection because of its prevalence in industry. This is supported

by results of a 1998 study performed by Venture Development Corporation which

indicated that Profibus DP was among the top three industrial communications buses

in popularity and usage [62].

Control strategies and mechanical modeling methods for web processes vary based

on the web characteristics. Strategies for controlling web tension using observers are

presented in [59] and [60], but are often unique to a process and involve tension state

calculation beyond the abilities of a typical PLC. Ultimately, spring-damper dynamic

models similar to those described in [63] and [64] were chosen as the modeling method

for the textile mechanics because they require fewer constants and calculation to

determine web tension in a PLC environment. Control of the textile tension control

16

PROFIBUS-DRIVE

INTERFACENODE NO.6

PROFIBUS A/DINPUT

MODULENODE NO. 4

PROFIBUS-DRIVE

INTERFACENODE NO. 5

SIEMENS 1HP208VAC3 PHASE

ADJUSTABLESPEED DRIVE

SIEMENS 1HP208VAC3 PHASE

ADJUSTABLESPEED DRIVE

SERIALINTERFACETO PLC CPU

PC RUNNING LABVIEWOPERATOR INTERFACE

PROGRAM TAKEUPFULL

SPOOLSENSOR

PAYOFFEMPTYSPOOL

SENSOR

ASI POWERCONVERTER

24VDC TO 30VDC

ASI BUSINPUTBLOCK

PC-PROFIBUS INTERFACECARD

PC RUNNING SCOPEPROFIBUS SOFTWARE

PC-PROFIBUS INTERFACECARD

ASI MASTERMODULE

PLC CPU /PROFIBUSMASTERMODULE

NODE NO. 3

ASI & A/DPOW ERSUPPLY:

24VDC OUT120VACSINGLE

PHASE IN

PLCMASTERPOW ERSUPPLY:

24VDC OUT120VACSINGLE

PHASE IN

PROFIBUS DATA BUS

PRIMARY ANALOGSENSOR INPUTS:

0-10VDC

ASI BUS

PLC RACK

1 HPINDUCTION

MOTOR:PAYOFF

1 HPINDUCTION

MOTOR:TAKEUP

SERIAL CABLE

Figure 2.2 Equipment configuration of experimental textile tension control system.

system is achieved by using two proportional-integral (PI) controllers with additive

state gain terms to compensate for variations in linespeed and tension.

Mathematical Subsystem Models

The tension controller signal flow diagram is shown in Figure 2.3. The mechanical

system is represented by a linear time-invariant state space system, with inputs of

motor torque, and outputs of linespeed and web tension. The state space plant model

derivation is provided in Appendix B.

17

A

B C

D

∫ +++

+

MECHANICAL PLANT

x.

xu

ydq MOTORRESPONSE

FREQUENCY/VOLTAGE

RESPONSE

ADJUSTABLE SPEEDDRIVES AND MOTORS

DESIREDLINESPEED AND

TENSION

CONTROLALGORITHM

(PLC)

FEEDBACKINSTRUMENTRESPONSES

BACKUP INSTRUMENTSIGNALS

A/DCONVERSION

RESPONSE

Figure 2.3 Tension control system block diagram.

Feedback instruments under normal operating conditions are modeled as linear

first order systems, where the measured mechanical value serves as an input, and an

analog 0-10VDC signal the output. The A/D converter block is also modeled in this

way, except the output is a signed 16-bit word, and the input an instrument signal

voltage. In its generic form, the first order relationship is described by

τd

dty(t) + y(t) = Kx(t) +B, (2.1)

where x(t) is the system input, y(t) is the system output, K is a gain constant, B a

constant output offset, and τ a time constant. To determine a first order subsystem

output, we solve for the system output y(t) in terms of a sum of integrals. Rearranging

Equation 2.1 gives usd

dty(t) =

K

τx(t)− 1

τ(y(t)−B), (2.2)

and integrating both sides yields

y(t) =K

τ

t

0x(t) dt− 1

τ

t

0(y(t)−B) dt. (2.3)

18

Equation 2.3 may then be used to determine y(t), provided that the initial conditions

of the integrals are known.

The variable speed AC drives are modeled as systems that deliver a three phase

sinusoidal voltage to the induction motors. The issued frequency, fe, varies as a first

order system with a signed word reference input and an electrical frequency output.

The AC drives operate using a constant V/Hz control scheme, which decreases the

output voltage magnitude with decreases in output frequency. At very low frequen-

cies, the voltage is increased slightly from the constant V/Hz line, and at zero Hertz

output frequency, the voltage is held above zero Volts at a constant level. This entire

scheme creates a frequency-RMS voltage relationship which may be described by

Vout = Vebaseemefe, (fe < 6 Hz) (2.4)

Vout = mlfe + Voffset, (fe ≥ 6 Hz) (2.5)

where fe is the motor output frequency, me is the exponential multiplier for low

frequency operation, Vebase is the zero Hertz base RMS voltage, ml is the linear

segment V/Hz slope, Voffset is the linear segment V/Hz offset, and Vout the RMS

line-line motor output voltage.

The motors are modeled using direct-quadrature (dq) motor theory [65], where

the motor inputs are sinusoidal three phase voltages, and the output is the developed

motor torque. To determine the developed motor torque, we calculate the dq voltages

from the drive output frequency and RMS line-line voltage magnitudes. The first step

is to determine the instantaneous three phase stator voltages using

vas =√2Vout cos(θ), (2.6)

vbs =√2Vout cos(θ − 2π

3), (2.7)

and

vcs =√2Vout cos(θ +

2π

3), (2.8)

where θ = ωet. These voltages are then used to calculate the stationary dq voltages

with

vsqs =2

3vas − 1

3vbs − 1

3vcs, (2.9)

19

and

vsds = −1√3vbs +

1√3vcs. (2.10)

The stationary voltages are then transformed into the synchronously rotating frame

voltages by

vqs = vsqs cos(θ)− vsds sin(θ), (2.11)

and

vds = vsqs sin(θ) + v

sds cos(θ). (2.12)

Additionally, for a shorted squirrel-cage induction motor the dq rotor voltages are

vqr = vdr = 0. (2.13)

The motor flux relationships in the dq model are

vqs = Rsiqs +d

dtΨqs + ωeΨds, (2.14)

vds = Rsids +d

dtΨds − ωeΨqs, (2.15)

vqr = Rriqr +d

dtΨqr + (ωe − ωr)Ψdr, (2.16)

and

vdr = Rridr +d

dtΨdr − (ωe − ωr)Ψqr. (2.17)

Equations 2.14-2.17 may be rearranged and integrated to solve for the dq fluxes, which

gives

Ψqs =t

0vqs dt−

t

0Rsiqs dt−

t

0ωeΨds dt, (2.18)

Ψds =t

0vds dt−

t

0Rsids dt−

t

0ωeΨqs dt, (2.19)

Ψqr =t

0vqr dt−

t

0Rriqr dt−

t

0(ωe − ωr)Ψdr dt, (2.20)

and

Ψdr =t

0vdr dt−

t

0Rridr dt−

t

0(ωe − ωr)Ψqr dt, (2.21)

20

which again may be easily calculated, provided that the initial conditions are known.

Having determined the flux components, the flux-current transform⎡⎢⎢⎢⎣iqsidsiqridr

⎤⎥⎥⎥⎦ =⎡⎢⎢⎢⎣(Lls + Lm) 0 Lm 0

0 (Lls + Lm) 0 LmLm 0 (Lls + Lm) 00 Lm 0 (Lls + Lm)

⎤⎥⎥⎥⎦−1 ⎡⎢⎢⎢⎣

Ψqs

Ψds

Ψqr

Ψdr

⎤⎥⎥⎥⎦ (2.22)

is used to calculate the dq currents in the machine. From these currents, the developed

torque is calculated by

Te =3

2

P

2Lm(iqsidr − idsiqr), (2.23)

where P is the number of poles in the machine. For a detailed derivation of dq motor

theory, the reader is referred to [65].

Figure 2.4 illustrates the control algorithm that the PLC executes. The linespeed

controller is essentially a takeup reel speed controller. The desired linespeed is divided

by the takeup reel radius to obtain the desired takeup reel speed. The actual takeup

reel speed is then subtracted from the desired value to create an error signal (SP-PV

operation). The error signal is then fed into a proportional-integral (PI) controller.

The load torque caused by tension is calculated using the takeup reel radius and

the tension feedback signal. This is multiplied by a gain constant and added to the

PI controller output as a tension compensation term. Finally, a gain intrinsic to PI

control in the Siemens PLC is applied, and the reference motor frequency signal is

sent to the takeup motor drive.

The tension control loop behaves similarly. The desired tension is multiplied by the

payoff reel radius to calculate a desired payoff tension torque. This is subtracted from

the actual payoff tension torque to create an error signal (PV-SP operation), which

is delivered to another PI controller. A state gain term to compensate for linespeed

variations is added to the output of the PI controller, which is then multiplied by the

intrinsic PI control gain and sent to the payoff drive.

Extensive testing of the textile tension controller was essential when determining

the constants used in the plant and instrument mathematical models. Appendix C

describes test procedures and provides tables of the determined values.

21

DESIREDLINESPEED 1/r1

+

-

θ1 FEEDBACK.

PI CONTROL#1 +

+

TENSIONFEEDBACK

SCALE TONEWTONS

TENSIONSTATE GAINr1

INTRINSICGAIN

TAKEUPREFERENCE(ROLLER #1)

DESIREDTENSION r5

-

+

++

SCALE TORAD/SEC

LINESPEEDSTATE GAIN

INTRINSICGAIN

PAYOFFREFERENCE(ROLLER #5)

θ5 FEEDBACK.

SCALE TORAD/SEC

TENSIONFEEDBACK r5

SCALE TONEWTONS

PI CONTROL#2

Figure 2.4 Controller for textile tension control system.

Simulation of Test Stand Behavior

The environment chosen for tension controller simulation was a line code imple-

mentation in Matlab. The use of line code was preferred over the graphical Matlab

Simulink environment because the system’s complexity makes a Simulink implemen-

tation unnecessarily complex and unwieldy. Furthermore, complete control over a

model’s behavior is more attainable in a line code environment. In a Simulink imple-

mentation, the ability to manipulate many built-in function blocks is limited.

General Overview

A flowchart for the code simulation is shown in Figure 2.5. A large main program

with no use of subfunctions was favored over many small subfunctions under a main

control program. This approach requires no variables to be passed between functions

and maintains all variable assignments in memory during program execution. The

duration of simulation was chosen to be a maximum of three seconds, which provides

adequate time to analyze prefault, faulted, and postfault conditions. For limited

22

‘staircasing’ of system signals, a resolution of 10,000 points per second was chosen,

or one evaluation point every 100µs. Altogether, the simulation may calculate 30,000

data points for each signal. Arrays of every intermediate (non-output) variable are

stored during simulation to help facilitate signal analysis throughout the system.

Variable Initialization

Setting the initial conditions of all variables is essential for establishing the starting

points of system signals. This requires both forward and backward calculation of

variables based upon the initial conditions of linespeed and tension.

The mechanical system initially requires all of the states in x and x to be defined.

These consist of values for all θi, θi, and θi. Given that the linespeed is constant at

t = 0, we calculate the values of θi and θi as

θi =vt=0ri, (2.24)

and

θi = 0. (2.25)

Calculations for the initial values of θi requires arrangement of Equations B.25, B.8,

B.9, and B.10 in matrix form, where the values of θi are unknown quantities. To

square the constants matrix, the value of θ3 is set to 10π (a value that would not lead

to a negative initial value for θ1). This arrangement creates the matrix equation⎡⎢⎢⎢⎢⎢⎢⎣2fp,t=0B2θ2B3θ3B4θ410π

⎤⎥⎥⎥⎥⎥⎥⎦ =

⎡⎢⎢⎢⎢⎢⎢⎣0 r2K23 −r3K23 + r3K34

r1r2K12 −r22K12 − r22K23 r2r3K23

0 r2r3K23 −r23K23 − r23K34

0 0 r3r4K34

0 0 1

· · ·

· · ·

−r4K34 00 0

r3r4K34 0−r24K34 − r24K45 r4r5K45

0 0

⎤⎥⎥⎥⎥⎥⎥⎦

⎡⎢⎢⎢⎢⎢⎢⎣θ1θ2θ3θ4θ5

⎤⎥⎥⎥⎥⎥⎥⎦ , (2.26)

where the solution for matrix θ is found by multiplying the inverse of the constants

matrix by the left hand side of Equation 2.26. Once values for x and x were deter-

mined, the input torques, u are determined using Equation B.1.

23

CLEAR MEMORY

DEFINE TIME ARRAYAND TIME VARIABLES

DEFINE MEASUREDPHYSICAL CONSTANTS

CONSTRUCT STATESPACE MATRICES

DEFINE LINESPEED ANDTENSION SETPOINTS

DEFINE GAIN, TIMECONSTANT, AND

OFFSETS FOR FIRSTORDER SUBSYSTEMS

DEFINE AC MOTOR ANDDRIVE CONSTANTS

DETERMINE INITIALCONDITIONS FOR ALLSYSTEM SIGNALS AND

INTEGRALS

CALCULATEUNDISTUBED FEEDBACKINSTRUMENT OUTPUTS

CALCULATE SAGGEDFEEDBACK INSTRUMENT

OUTPUTS

CALCULATE A/D INPUTBLOCK OUTPUTS

CALCULATE CONTROLSIGNALS

CALCULATE DRIVESETPOINT SATURATION,OUTPUT FREQUENCY,DC BUS STATUS AND

OUTPUT VOLTAGE

CALCULATE DEVELOPEDMOTOR TORQUES

CALCULATEMECHANICAL STATE AND

OUTPUT VALUES

DEFINITIONS AND SETUP LOOP: INDIVIDUAL DATAPOINT EVALUTION

STARTSIMULATION

ENDSIMULATION

LAST TIME EVALUATIONPOINT REACHED?

YES

NO

LOAD STOREDDATA FOR

INSTRUMENT SAGDEVIATIONSEGMENT

DEFINE CONTROLCONSTANTS

Figure 2.5 Flowchart of tension controller voltage sag simulation program.

24

The instrument output initial conditions were determined by multiplying the phys-

ical input by the instrument gain, then adding the instrument offset. This operation

was also performed for the A/D inputs. Note that this is a steady state interpreta-

tion of Equation 2.1. The initial conditions of the integral sums used to calculate the

outputs of these devices were then assigned a value of the output at t = 0.

Calculation of the initial dq voltage, flux, and current conditions for the motors

could not be done in reverse using the output torques because several unique combi-

nations of dq currents could lead to the same developed torque. Therefore, for each

motor an iterative scan and forward calculation through multiple electrical frequency

setpoints was performed, and the resulting output torque compared to the actual ini-

tial torque. The input electrical frequency that yielded an output torque equal to the

actual initial output torque was then stored as the initial condition for the motor’s

desired electrical frequency. Calculation of initial line voltages were determined using

the V/Hz drive scheme described by Equations 2.4 and 2.5. The initial dq voltages

were calculated using Equations 2.6 through 2.13, and dq currents were calculated

assuming steady state operation by

⎡⎢⎢⎢⎣iqsidsiqridr

⎤⎥⎥⎥⎦ =⎡⎢⎢⎢⎣

Rs ωeLs 0 ωeLm−ωeLs Rs −ωeLm 00 (ωe − ωr)Lm Rr (ωe − ωr)Lr

−(ωe − ωr)Lm 0 −(ωe − ωr)Lr Rr

⎤⎥⎥⎥⎦−1 ⎡⎢⎢⎢⎣

vqsvdsvqrvdr

⎤⎥⎥⎥⎦ .(2.27)

Initial motor fluxes were then calculated using the flux-current transform in Equation

2.22, and the initial values of the corresponding integral sums were assigned the values

of the fluxes at t = 0.

Some control signals also required calculation of their initial conditions. The

outputs of the integral controllers were calculated assuming that the controller errors

at t = 0 are zero, and the proportional contributions to controller output are zero as

well. With these values set, the integral sums were back calculated using the initial

electrical frequency outputs and the initial feedback signal values.

25

Numeric Evaluation of Integrals

The first order solution in Equation 2.3 is obtained numerically, where the value of

the integrals at t = t−dt is calculated and multiplied by dt, then added to the storedintegral sum to yield the output value. Since the integrals are calculated using values

from the previous step in time, any error in this calculation can be decreased by using

smaller dt time step. This evaluation method was used for the feedback instruments,

A/D converter, first order drive setpoint transitions, and the dq flux calculations.

Data Sampling

The data sampling that occurs in the PLC, data bus, and I/O modules is modeled

by the inclusion of zero order holds in the simulation. This is achieved by referencing

signals at previous times which are determined by

tref = Tst

Ts(2.28)

where Ts is the zero order hold sampling period, and tref is the previous time reference.

Process Response to Voltage Sags

The textile tension control stand’s overall response to voltage sags is dependent

on the individual hardware sag responses. Because of this, the analysis of process

behavior must initially concentrate on the component devices.

Analog Instrumentation

Assessment of instrument voltage sag responses may take a pass/fail approach,

where device responses are described with terms such as ‘immune’ or ‘tolerant’ to volt-

age sags [66]. More clearly specified are voltage sag tolerances described by CBEMA

or ITI curves where magnitude and duration thresholds are defined in order to deter-

mine device compliance with a particular specification [46]. Meeting these standards

is open to interpretation for analog process sensors because during some sags the

device will continue to operate, but may not be considered as surviving the event due

to an erroneous output signal.

26

In 2000, a study performed by EPRI Solutions (formerly EPRI PEAC) examined

the voltage sag response characteristics of a wide range of process sensors. The

findings from this study offer some insight into general sensor behavior which we

aim to analyze. It was shown that sensor responses were not sensitive to the point-

on-wave characteristic of voltage sags, which suggests that point-on-wave variations

would be an unnecessary addition to an instrument sag response model. On the other

hand, the sensor output level at the time a voltage sag is applied must be taken into

consideration because the output level affects the device loading, which in turn affects

the ridethrough time [67].

Figure 2.6 shows the response of the payoff roller tachometer, takeup roller tachome-

ter, and tension sensor to a 0% sag (interruption) for a duration of 450ms. Our initial

goal is to model the output response of each sensor so that they may be implemented

in the Matlab process simulation routine. In the past, models of corrupted sensor out-

puts have focused on injecting a subtractive disturbance [68] or a variable gain [69]

to describe the initial decays seen in Figure 2.6. The shortcomings of these models,

however, are that erratic device discharge and gradual recovery are not accounted for.

The improved instrument sag response model aims at including these factors, while

viewing the instruments as ‘black boxes’, whose internal circuitry is of less importance

than the overall input-output characteristics.

To develop a generic analog instrument sag response model, we begin by examining

the interconnections that are most frequently used for these devices. Figure 2.7

shows common topologies for both AC and DC powered instruments. DC powered

instruments are typically fed by a separate AC to DC converting power supply, which

may also feed other DC loads. In this case, it is important to note that the presence of

additional loads will affect the sag response of the instrument because of the energy

demands these loads place on the supply. The physical input to the device also

affects the response because it places varying demands on the instrument’s stored

energy during the sag. AC supplied instruments behave similarly, only without the

presence of additional DC loads.

27

Figure 2.6 Textile tension control stand instrumentation voltage sag responses.

DC SUPPLIEDINSTRUMENT

DCPOWERSUPPLY

ANALOG SIGNALOUTPUT

OTHER DC LOADS

AC

INPUTVOLTAGE

(SAG MAGNITUDE,DURATION)

PHYSICALINPUT

AC SUPPLIEDINSTRUMENT

PHYSICALINPUT

ANALOG SIGNALOUTPUT

a.

b.

ACINPUT

VOLTAGE(SAG MAGNITUDE,

DURATION)

Figure 2.7 Instrument input-output topologies for a) DC sensor fed through separatepower supply and b) AC supplied sensor.

28

Figure 2.8 illustrates the output response of several instruments when a 10%

voltage sag is applied for a duration of 450ms. Each response may be divided into

a series of operating segments. The prefault segment occurs before any disturbances

have been applied to the sensor. During this condition, the instrument behaves

normally. Similarly, the postfault segment occurs after the supply disturbance and all

of its effects have cleared, and is also treated as an undisturbed instrument response.

Figure 2.8 Instrument voltage sag responses divided into segments of operation.

Ridethrough Segment

The ridethrough segment of the voltage sag response occurs from the time of sag

inception to the time at which the output signal begins to deteriorate, and is consid-

ered a general indication of the energy stored in a device’s power supply. During this

segment the instrument functions normally. The duration of the ridethrough segment

29

is dependent on both sag magnitude and the physical input level, as illustrated in

Figure 2.9. The magnitude of the voltage sag has an effect on the ridethrough times

because different amounts of energy are drawn at various supply magnitudes, while

the physical input level places varying demands on the instrument’s stored energy.

When describing the ridethrough segment, sag duration is only a consideration

if it is less than the ridethrough time given a certain magnitude and physical input

level. If the sag duration is shorter, then the ridethrough time is equal to the sag

duration. Under these circumstances, the instrument may be considered immune to

a voltage sag, and the deviation and recovery segments are not present.

Figure 2.9 Effects of variations in sag magnitude and physical input level on ride-through segment of instrument voltage sag response.

30

Deviation Segment

The deviation segment of the instrument voltage sag response begins at the point

where the output signal begins to deviate from its prefault and ridethrough values.

It terminates at the time of sag extinction. Often this response takes on the form

of a decaying exponential signal, but it can also contain spikes and transients when

components in the instrument de-energize. Again, both the voltage sag magnitude

and physical input level have an effect on the nature of this response. Figure 2.10

exemplifies this for several cases. The duration of the voltage sag only defines the time

at which the segment ends, and does not affect the nature of the deviated waveform.

Figure 2.11 demonstrates the effects of duration changes given constant magnitude

and physical input conditions.

31

Figure 2.10 Effects of variations in sag magnitude and physical input level on deviationsegment of instrument voltage sag response.

Recovery Segment

The recovery segment begins at the time of sag extinction. Typically this involves

an exponential rise from the output at the time of sag extinction to the unfaulted

instrument output level. Previous models have treated this as a simple step function,

which implies that the output immediately returns to normal when the sag ends.

Figure 2.12 illustrates the exponential nature of device recovery, and that factors of

sag magnitude and physical input affect recovery times. Sag duration also has an

effect on the recovery time, since the duration plays a part in determining initial

conditions at the time of recovery.

32

Figure 2.11 Effects of variations in sag duration on deviation segment of instrumentvoltage sag response.

An important property of device recovery is that the output signal gradually

transitions from a corrupted signal to an uncorrupted one. When the device is fully

recovered, the output is expected to be an accurate representation of the physical

input. A mathematical description of this behavior must account for changing input

levels during the recovery process, rather than use a constant to describe the final

signal value. This may be modeled as

Output(t) = (Outputt=text)(e−(t−trec)

τrec ) + F (Input(t))(1− e−(t−trec)τrec ), (2.29)

where trec is the time of sag recovery, τrec the recovery rise time constant, F the

transfer function for the undisturbed instrument, and Input(t) the device physical

33

Figure 2.12 Effects of variations in sag magnitude and physical input level on recoverysegment of instrument voltage sag response.

input as a function of time. This representation of the recovery response is a sum of

the corrupted signal final value scaled to become less significant with time and the

uncorrupted signal scaled to become more significant with time. Figure 2.13 shows

a comparison of the theoretical output when applied to an actual recovery signal

assuming a linear undisturbed instrument response and constant physical input.

Implementing Results in Simulation

The observed ridethrough times, maximum recovery time, and deviation segment

trace data for a particular sag magnitude and device physical input combination may

be applied in a simulation environment. To implement in the Matlab simulation of

the textile tension control stand, an if-then structure is used to output each segment

34

Figure 2.13 Application of mathematical recovery response model to instrument usingconstant physical input level.

response at a specific time. Figure 2.14 shows a flowchart for instrument sag response

simulation and Table 2.1 describes the event schedule that controls the segment tran-

sitions. For prefault conditions, the normal instrument output is delivered to the

remainder of the control system. The same is true for the ridethrough segment, up to

the point where t = tstart + tridethru. This point marks the beginning of the deviation

segment.

Because of the variety of deviated waveforms that may be produced by a particu-

lar instrument during the deviation segment, the program references the appropriate

output to deliver from a sampled waveform obtained during sag testing. This avoids

the impractical task of calculating output values based on a customized instrument

response model, and meets the requirement of being able to simulate a value that cor-

35

Table 2.1 Event schedule for implementation of improved instrument voltage sagresponse model in simulation.

SEGMENT TIME FRAMEPrefault tstart < t < tincepRidethrough tincep < t < (tincep + tridethru)Deviation (tincep + tridethru) < t < trecRecovery trec < t < (text + 10τridethru)Postfault (trec + 10τridethru) < t < tend

rectly represents an erratic output waveform. In this step, it is important to maintain

consistency between sample times in simulation and in sag testing. Furthermore, it

is preferable to use a high sampling frequency to avoid waveform ‘staircasing’ during

the deviation segment simulation. The duration of the theoretical sag may be varied

by switching from the sampled waveform in the deviation segment to the recovery

segment before the entire sampled waveform is sent as an output.

When the sag has ended at trec, the program outputs an implementation of Equa-

tion 2.29. Here the value of τrec, the recovery time constant, is equal to one fifth of

the maximum recovery time measured for the device. The equation is calculated and

delivered as a device output until ten recovery time constants have elapsed following

text, at which point the normal operating conditions are again restored as the device

transfer function. The process simulation Matlab code, including the instrument sag

response models, are included in Appendix D.

Results for instrumentation voltage sag tests are shown in Appendix E. Each

sensor was tested to determine ridethrough times at varying physical input levels.

The physical levels used were taken over a range from the lowest allowable process

physical input level to the highest, and did not use the output limits of 0-10VDC

as the test limits. Sag magnitudes ranged from 0% to 80% in 20% increments and

durations were adjusted to determine the maximum recovery time.

AC Motor Drives

Both of the AC motor drives operate using rectifier/inverter topology. The main

effect that voltage sags have on the AC motor drives is to lower the DC bus magnitude

36

START INSTRUMENT SAGRESPONSE SIMULATION

CALCULATEUNDISTURBED

INSTRUMENT RESPONSE

SIMULATION INRIDETHROUGH SEGMENT?

YES

NO

END INSTRUMENT SAGRESPONSE SIMULATION

SIMULATION IN PREFAULT ORPOSTFAULT OPERATION?

SIMULATION INDEVIATION SEGMENT?

NO

YES

NO

OUTPUTUNDISTURBEDINSTRUMENTRESPONSE

OUTPUTRECORDED

INSTRUMENTRESPONSE

DATA

OUTPUTRECOVERY

MODELRESPONSE

SIMULATION INRECOVERY SEGMENT:CALCULATE OUTPUT

ACCORDING TORECOVERY MODEL

YES

Figure 2.14 Flowchart for simulation of instrument voltage sag response.

37

between the rectifier and controlled inverter below their nominal prefault levels. Both

the magnitude and phasing combination of the incoming voltage sags have an effect

on the DC bus level. Figure 2.15 shows the DC bus response for two different events.

The left column shows a 60%, three phase voltage sag of 450ms duration and the

corresponding DC bus response. For this sag, the DC bus level drops below the

trip level of approximately 200VDC, where the drive halts its output to the motor.

Afterward, the DC bus magnitude continues to decline, but at a lower rate because

energy is no longer consumed by the running motor. After an output trip and sag

recovery, the DC bus returns to its nominal magnitude, but experiences a brief period

of ripple that matches the peaks of the incoming sinusoidal voltages.

The right column of Figure 2.15 shows the DC bus magnitude of the same drive

when it is stricken by an 80%, three phase voltage sag of 450ms duration. In this

case, the DC bus magnitude drops sharply after the sag inception, but remains above

the 200VDC trip level. The motor does not trip offline, and the DC bus does not

ripple after sag recovery.

38

Figure 2.15 DC bus levels during three phase sags of magnitude 60% (left) and 80%(right).

For unbalanced sags, the response is similar. Maintenance of the DC bus above

the trip level defines its ability to ride through the event, and bus decay occurs in the

same fashion as in the balanced case. For sag magnitude and phasing combinations

that cause an output trip, significant DC bus ripple can occur after the dropout. For

combinations that do not cause a dropout, increased ripple can occur during the sag

when the DC bus has reached its steady state depressed level.

Figure 2.15 illustrated the difference in the DC bus magnitude for both trip and

ridethrough conditions, but assumed a sag duration long enough for the magnitude

and phasing combination to be the most significant factors in determining the drive’s

39

ability to withstand the sag. Figure 2.16 shows the effects that shortened durations

can have on the drive DC bus magnitude. The left column shows the response to

varying durations for a sag magnitude/phase combination that causes an output trip,

and shows the rapid decay of the DC bus under these conditions. During the sag,

the drive briefly continues to run the motor, but quickly reaches the dropout voltage

and halts the output. This suggests that while the duration of the incoming sag is a

factor in determining drive tolerance, it is only critical during the first few cycles of

the sag. The right column of Figure 2.16 shows the effects of varying sag duration

for a magnitude/phasing combination that does not cause a drive output trip. Here

these effects are even less significant, as sag duration only determines the length of

time that the DC bus stays at a depressed level.

The rate of decent of the DC bus is also a variable characteristic, and is dependent

on motor loading. Figure 2.17 demonstrates this effect. For the sag shown in the top

waveform, the DC bus is shown for various loading conditions. The second waveform

shows the steep decent in the DC bus during heavily loaded conditions, and the third

waveform shows a more gradual decay when the motor is lightly loaded. The last

waveform shows the no load rate of DC bus decay, which was measured when the

motor output was de-energized. During a sag magnitude and phasing combination

that will ultimately cause a drive trip, the DC bus magnitude is clearly dependent

on the motor load, and decays at a far slower rate when the motor output is off

altogether.

Several variables have been shown to affect the ridethrough characteristics of the

textile tension control stand’s AC motor drives. While motor loading and sag duration

are factors, they are overshadowed by the influence of sag magnitude and phasing

combinations. In the case of the experimental process, dropout characteristics were

tested for a range of magnitude and phasing combinations. These results are listed

in Appendix F. Each test was conducted by delivering a 450ms sag of a certain

magnitude and phasing combination. During the tests, the motors were energized

and running in the textile process. The recorded response is the drive’s ability to ride

through the event without dropping out and requiring a user restart.

40

Figure 2.16 Sag duration effects on the motor drive DC bus levels.

41

Figure 2.17 Loading effects on the motor drive DC bus decay rates.

Simulation of these effects are implemented in the main Matlab simulation pro-

gram by forcing drive output voltages to zero after calculations for the DC bus levels

indicate that the dropout threshold has been crossed. In the simulation, the measur-

able sag characteristics are not considered. Instead, the observed DC bus decay rate

is inserted directly into the routine to simulate a sag that causes drive dropout.

Additional Immune Hardware

Additional hardware present in the textile tension control stand (see Figure 2.2)

was tested with a sag generator for a 450ms, 0% sag (interruption) and rode through

the test events without disturbance. These devices are listed in Table 2.2. Addition-

42

ally, all AC contactors present in the system were bypassed for system-wide voltage

sag tests, since their ridethrough characteristics can be mitigated with inexpensive

hold-in devices.

Table 2.2 Tension stand hardware that showed no dropout response.

DEVICE NOTESSiemens PLC power supply 24VDC, feeds PLC onlySiemens PLC fed by dedicated supply5A, 24VDC supply supplies I/O block, AS-I converter, limit sensorsAS-I network power converter 24VDC to 30VDC converterAS-I network master fed by 5A, 24VDC supply through converterProfibus I/O block fed by 5A, 24VDC supply

Combined Process

Once individual device voltage sag responses are understood, the integrated process

is examined as a whole. If we initially consider the process as one entity, then the volt-

age sag tolerance may be visualized as shown in Figure 2.18. Each axis of this graph

represents the line-neutral voltage magnitude of an incoming power supply phase.

Voltage sags that fall within the shaded region cause the process to malfunction,

while sags that fall outside the region have no effect.

All sags that were delivered to the textile tension controller to generate Figure 2.18

were 450ms in duration, which allowed ample time for the AC motor drives to trip

offline, and for the instrumentation to deviate from their prefault values significantly.

The boundary between regions falls at the midpoint between a sag test magnitude

point which yielded a disruption and one that did not. Figure 2.18 therefore represents

the 450ms process disturbance/no disturbance test results with 20% sag magnitude

increments, for a total of 216 tests.

The response may be further divided into regions of specific malfunction. The

most catastrophic of these regions is shown in Figure 2.19. This region indicates the

sag magnitude and phasing combinations where the AC drives drop offline, which

constitutes a complete process shutdown. This type of failure requires user input to

restart the process and manually rethread the machine.

43

Figure 2.18 Region of voltage sags for any process disturbance in textile tensioncontrol system.

44

Figure 2.19 Region of voltage sags for AC motor drive dropout in textile tensioncontrol system.

45

The AC drive dropout region boundaries may be described mathematically. Figure

2.20 shows a scatter plot of AC drive voltage sag dropout points in two phases. Here

phase C is held at a constant 0% magnitude, while phases A and B are variable.

Using the magnitude midpoints between tests that caused a dropout and those that

rode through, a threshold is established with the line

A+B = 1.5, (2.30)

where A and B are the per unit phase voltage magnitudes. This threshold is similar

for plots of phases B versus C and C versus A with the missing phase voltage held

at zero. Three inequality statements result from these boundary lines. The logical

AND of the inequalities form an expression to indicate a sag’s presence in the AC

drive dropout region. This may be expressed by

Q1 = (A+B < 1.5) AND (B + C < 1.5) AND (C +A < 1.5), (2.31)

where Q1 is a Boolean term indicating sag presence in the AC drive dropout region.

A comparison of actual and simulated behavior for this case is shown in Figure

2.21. An immediate loss of tension occurs when the drives drop offline, and the process

linespeed slowly decelerates. Only slight differences exist between the experimental

and simulated waveforms. The experimentally recorded tension waveform shows a

slightly slower rate of decent to the zero tension state than its simulated counterpart,

which may be attributed to the first order delay of the tension cell. In the simulated

linespeed waveform, the decay experiences a momentary reversal that is not seen in

the measured linespeed. This is caused by the mechanical modeling of the rollers

remaining coupled during the zero tension state, while the actual system experiences

decoupled rollers with web slack between them.

Although the AC drive dropout region accounts for the majority of possible process

malfunctions, another region exists that requires attention. This malfunction region

is shown in Figure 2.22, and represents the voltage sags that cause process variables

to go out of tolerance. In this region, the AC drives do not trip offline due to a DC

bus undervoltage. Instead, the instrumentation delivers erroneous process feedback

46

Figure 2.20 Two phase graph of AC drive dropout points and dropout boundaryestimation line.

47

Figure 2.21 Actual and simulated textile tension controller responses during motordrive dropout.

48

Figure 2.22 Region of voltage sags for instrumentation malfunction in textile tensioncontrol system.

49

information to the controller. This in turn causes the controller to call for a change

in the process that is not required, which results in a deviation from desired process

values. Some sags that affect the instrumentation are outside the indicated region, but

still exist within the AC drive dropout region. They are therefore neglected because

AC drive dropout is the more dominant, catastrophic variety of process malfunction.

Sag presence in the faulted instrumentation region may also be described in math-

ematical terms. By inspection of Figure 2.22, we observe that the faulted instrumen-

tation region is present where the Phase C magnitude is below 70%, and the sag

magnitude and phasing combination falls outside the AC drive dropout region. This

is expressed as

Q2 = (Q1 = FALSE) AND (C < 0.7), (2.32)

where Q2 is a Boolean value indicating a sag’s presence in the faulted instrumentation

region.

Simulation of an instrumentation level malfunction is compared with measured

values in Figure 2.23 using the Matlab simulation. A single phase, 0% voltage sag

(interruption) is delivered to Phase C, and both the tension cell and tachometers

are driven into a faulted state. Their feedback values are misinterpreted as physical

disturbances, and compensated for by the PLC control algorithm. The actual web

tension is driven to over double its prefault value, and the linespeed sees a significant

disturbance as well. The ridethrough, deviation, and recovery segments in the in-

strument responses are all clearly visible, and their theoretical models follow closely.

Discrepancies between the measured and simulated responses can be attributed to

slight differences in measured component gains, unmodeled roller imbalance effects,

and noise.

Summary

A network connected textile tension control system was constructed with the

intended purpose of serving as a test bed for developing software based voltage sag

mitigation algorithms. After mathematical analysis and detailed measurement of

50

Figure 2.23 Comparison of actual and simulated textile tension controller response to450ms, 0% sag (interruption) affecting the instrumentation only.

51

constants, a Matlab based program was written to simulate the tension controller’s

dynamic behavior.

During analysis of individual device behavior, it was found that an improved

instrumentation voltage sag response model was required to effectively predict the

process voltage sag output response. This model improves upon previous mathemati-

cal instrument response models by accounting for erratic device output and modeling

gradual instrument recovery characteristics. The developed model was implemented

in the Matlab simulation. AC motor drive voltage sag responses were found to fol-

low a ride through or trip dichotomy, the result of which is highly dependent on sag

magnitude and phase combinations. Other devices such as DC power supplies were

found to have no effect on the process voltage sag response due to their preexisting

ridethrough and light loading.

Using a series of voltage sag tests at 450ms durations, varying forms of the com-

bined process response were observed. These consist of three distinct regions:

1. an AC drive dropout region, which was found to be the most catastrophic

because of the required user input to remedy;

2. a faulted instrumentation region, which causes corrupted feedback signals to

force the existing control algorithm to call for unnecessary corrective action;

3. a no disruption region, where the process behaves normally and rides through

the voltage sag.

Simulations of the AC drive dropout region and the faulted instrumentation region

were performed and matched with experimental results.

Using the textile tension control stand as a research test bed, our focus now turns

to the software architecture and algorithm development used to mitigate the voltage

sag responses observed in this system.

52

CHAPTER 3

SOFTWARE DESIGN

Introduction

A coordinated voltage sag mitigation software suite was designed and implemented

for use with the textile tension control stand. This program aims at minimizing the

addition of hardware and being unintrusive to existing process controllers.

The software resides on a standalone PC, termed ‘Ridethrough PC’, and interfaces

with both the existing Profibus automation network and the process main three-phase

power supply. The Ridethrough PC executes a single program which continually

monitors the incoming power supply, detects voltage sags, performs fast sag parameter

measurements, determines an expected process response, and executes a mitigation

routine based on user input and the expected process voltage sag response. Historical

process data can also be stored in the Ridethrough PC. Process control changes are

executed in the existing PLC, but are called upon by override commands originating

from the Ridethrough PC.

Communications and Interfacing

Figure 3.1 shows the general interfacing topology of the Ridethrough PC. Inter-

facing details are described in terms of its component subsystems.

PC Specifications

The Ridethrough PC contains a 2.66 GHz Pentium 4 processor, 512 MB of RAM,

and runs the Windows XP Professional operating system. LabView version 6.1 is

the main software environment for programming and execution of the detection and

mitigation routines. LabView was chosen because of its data acquisition interface and

analysis capabilities. During the software development process, LabView’s graphical

programming environment allowed for extensive variable monitoring and analysis,

thereby decreasing troubleshooting time.

53

PROFIBUSNETWORK PROFIBUS

INTERFACE CARD(SLAVE CONFIG.)

COMMAND ANDCONTROL DLL

FUNCTIONS

RIDETHROUGH PC

LABVIEW

DATA ACQUISITIONINTERFACE(LABVIEW)

DATA ACQUISITIONCARD (PCI-1200)

VOLTAGEDIVISION AND

ISOLATION

THREE PHASE SUPPLY(SAG GENERATOR)

TEXTILE TENSIONCONTROL STAND

Figure 3.1 Ridethrough PC interface structure.

Profibus Interface

The Profibus interface card used in the Ridethrough PC is a Woodhead/SST

automation network interface card, model SST-PFB3-PCI. It is capable of performing

as both a Profibus network master or a network slave device. In the case of the

Ridethrough PC, it is programmed as a slave device with 16 bytes of transferred

input data and 16 bytes of output data. It can function at Profibus baud rates of up

to 12Mbps, but is set to operate at the process network’s baud rate of 1.5Mbps.

Function calls to the Profibus interface card were performed in LabView by calling

individual functions in a dynamic link library (.dll) file. Upon startup, the LabView

program calls a subroutine that opens the card for communications, defines network

and node parameters, and switches to an online state. During the mitigation pro-

gram, read and write functions are performed during every scan cycle of the monitor

program. The speed at which data is passed to the PLC from the LabView program is

therefore defined by the scan cycle time of the ridethrough program and the Profibus

network scan cycle time. When execution of the ridethrough program is terminated,

54

a sequence to switch the card offline from the Profibus network is executed. LabView

code for Profibus card initialization, I/O, and shutdown is provided in Appendix G.

PLC Communications

The PLC received basic programming modifications to accommodate an additional

node on the Profibus network. The additional Ridethrough PC slave node was added

to the network using the .gsd messaging configuration file provided by the Profibus

card manufacturer. A user-defined setting of 16 input bytes and 16 output bytes was

defined in the network configuration environment.

Addition of another node in the Profibus slave scan list added approximately

650µs to the network scan time, and averages 5.9ms per scan with the additional

node. Memory requirements of the Ridethrough PC did not constitute an appreciable

change in PLC memory usage, with the existing controller utilizing only 10% of the

available 98,304 bytes of memory space.

Data Acquisition

A National Instruments PCI-1200 data acquisition card was used for power supply

signal monitoring. This card allows up to eight analog inputs (0-10V or ±5V), andsamples inputs with 12 bit resolution at a maximum rate of 100kS/s. The Ridethrough

PC uses four of these analog input channels. Three are for each supply voltage, and

the fourth is used for a trigger signal during testing. Sampling rates are set at 6kHz

for each channel, which yields 100 data points per cycle of 60Hz voltage. Analog and

digital outputs are also available with the PCI-1200.

Initialization and access to the data acquisition card is accomplished through

functions provided with LabView. Parameters of device identification, desired input

channels, and sampling rates are user defined inputs to these functions. Outputs of

these functions are raw data, actual sampling times, and scan buffer backlog levels.

55

Power Supply Interface

The data acquisition card is connected to the main three phase power supply

through separate voltage dividers that output 9.24V peak-to-peak for a 120VAC RMS

input. The signals are then isolated with a Gould voltage isolation amplifier set to a

1:1 attenuation. These signals are then fed directly into the data acquisition card.

User Interface

After calling initialization functions for the Profibus and data acquisition cards,

the ridethrough program executes a user-terminated while loop in which user com-

mands may dictate the following:

1. Read historical process response information into memory.

2. Write recorded response information in a data file after program termination

(for use in subsequent program executions).

3. Associate incoming sags with those recorded during process data accumulation.

4. Calculate the process expected voltage sag response.

5. Execute ridethrough algorithms based on user algorithm selection and the ex-

pected process response.

Indicators of sag magnitude and phasing are available through the user interface.

The graphical environment also allows for a ‘virtual oscilloscope’ display of incoming

voltage waveforms. The user interface is shown in Figure 3.2.

56

Figure 3.2 Screenshot of Ridethrough PC user interface.

Existing System Modifications

Changes to the existing system consist of an additional subroutine inserted at the

end of the main PLC ladder program, and a series of switching commands included

in the main program. The existing PLC program is modified to continuously output

critical process data and node status information to the Ridethrough PC, and also

accept alternate control commands dictated by the Ridethrough PC during and after

a voltage sag.

Signal Switching

Under normal operation, the main PLC program transfers process data from the

Profibus input buffer to the PLC memory area. Interrupt switches were inserted in

all of the main signal transfer commands, which is achieved with a ‘normally closed

contact’ operation. This enables the move command to execute if the transfer bit

is low. Transfer bits are defined in the Ridethrough PC control word, and their

57

corresponding signals are listed in Table 3.1. The main PLC ladder logic code is

listed in Appendix H.

Also included in the control word is a drive coast command, which sends motor

drive output disable commands on a rising edge and drive restart commands on a

falling edge.

Table 3.1 Ridethrough PC control word structure.

BIT(S) COMMAND0 Drive coast command1-7 unused8 Replace bit - tension value9 Replace bit - payoff tachometer value10 Replace bit - takeup tachometer value11 Replace bit - payoff drive output12 Replace bit - takeup drive output12 Replace bit - linespeed setpoint14 Replace bit - tension setpoint15 unused

PLC Subroutine

The added PLC subroutine transfers raw process data to the Ridethrough PC

for analysis and storage along with a word for motor drive and I/O node status

data (Table 3.2). Ladder code for the subroutine is listed in Appendix H, while

functionality is described in the flowchart in Figure 3.3. The subroutine is called

during every PLC scan cycle, and adds approximately 4ms to the total program

execution time.

Table 3.2 Ridethrough PC status word structure.

BIT(S) STATUS0 Payoff drive active1 Takeup drive active2 I/O block active3-15 unused

58

START PLCSUBROUTINE

MOVE INSTRUMENT SIGNALS,DRIVE REFERENCE SIGNALS,

STATUS INFORMATIONTO DATA BUS

FOR ALL REPLACE BITS:IS REPLACE BIT i TRUE?

YES

NO

MOVE REPLACEMENTSIGNAL i INTO CONTROL

LOOP

DRIVE HALT COMMANDFROM RIDETHROUGH PC

RISING?

YES

NO

DRIVE ENABLE BITS =FALSE

DRIVE HALT COMMANDFROM RIDETHROUGH PC

FALLING?

YES DRIVE ENABLE BITS = TRUE,DRIVE START BITS = TRUE

ANY DRIVE REPLACE BIT = TRUE ORDRIVE COAST COMMAND = TRUE

YES

NO

CONTROL INTEGRATORHOLD = TRUE

SEND REPLACEMENTREFERENCE SIGNALS TO

DRIVES

FOR ALL BUS NODES:NODE i ACTIVE?

YES

NO

NODE ACTIVE BIT i =TRUE

END PLC SUBROUTINE(RETURN TO MAIN)

NO

Figure 3.3 Flowchart of added PLC subroutine.

59

Additional Commands

Additional commands introduced into the PLC program were minimal. Integral

control hold bits were added to the PI control blocks for the linespeed and tension

controllers. Provisions for these commands already existed in the PI control blocks,

but were unused before the addition of the Ridethrough PC. Lastly, the Ridethrough

PC subroutine call command was added. This executes during every program cycle

of the main PLC program

Voltage Sag Detection

Figure 3.4 shows a flowchart for the method of voltage sag detection and measure-

ment. The Ridethrough PC accepts three phase voltage signals and samples them

at a rate of 100 sample points per input voltage cycle (6kHz). The sampled data

is delivered to the main program and subjected to a sag detection algorithm which

generates a Boolean detect signal for each phase. Detection is performed by analysis

of RMS voltage derivatives. Sag detect signals are then used by the main program to

initiate data recording or process mitigation routines.

For each while loop cycle of the ridethrough program, five voltage data points are

sampled from each phase. After scaling the input signals to compensate for the input

voltage dividers, the data points are inserted into an existing voltage data array of 100

elements, where the oldest five samples are discarded and the newly sampled points

are shift-inserted. After the discard and insert operation, the 100 element array is

used to calculate the present RMS voltage with

VRMSk =1

n

n−1

i=0

v2i , (3.1)

where the array size n = 100, k is the cycle index for the ridethrough program, and

vi the voltage sample data points. The RMS derivative is then calculated using

d

dtVRMSk =

VRMSk − VRMSk−1dt

(3.2)

where dt is the time differential between Ridethrough PC cycles, or

dt =5

fsample= 0.833ms. (3.3)

60

READ FIVEVOLTAGE DATA

POINTS FOR EACHPHASE

SCALE INPUT DATATO ORIGINAL

VOLTAGE BASE

INSERT NEW DATAPOINTS INTO

VOLTAGE ARRAYS

DISCARD OLDESTFIVE DATA POINTS

FROM VOLTAGEARRAYS

CALCULATE RMSVOLTAGES - SINGLE

CYCLE WINDOW

CALCULATEMAGNITUDES USING

QUARTER CYCLEPEAK DETECT

CALCUATE RMSVOLTAGE

DERIVATIVES: dt=5Ts

START MEASUREAND DETECT

FOR EACH PHASE: (WAS_SAG =FALSE AND dVRMS < NEGATIVE

THRESHOLD) OR (WAS SAG = TRUEAND dVRMS > POSITIVE

THRESHOLD)?

YES

NO

IS_SAG PHASE i =INVERT (WAS_SAG

PHASE i)

ANY PHASE: IS_SAGPHASE i = TRUE?

YES

NO

IS_ANY_SAG =FALSE

IS_ANY_SAG =TRUE

END MEASUREAND DETECT

Figure 3.4 Flowchart of voltage sag measurement and detection routine.

61

In the prefault and postfault cases, the RMS derivative remains at zero. Sag inception

is detected when the RMS derivative crosses a negative threshold, and recovery is

detected when the RMS derivative crosses a positive threshold.

After initializing all detect signals to logic ‘false’, the detection algorithm performs

an invert operation to the detect signal when the first below-threshold RMS derivative

voltage occurs. Subsequent crossings of the negative threshold are ignored. The

detect signal is inverted back to the logic ‘false’ state upon the first above-threshold

RMS derivative voltage seen while in the ‘true’ state. Initial conditions are of critical

importance, as unpaired inception or recovery triggers could potentially lead to a false

detect signal.

Theoretical Response

The voltage sag algorithm was implemented in Matlab to simulate its performance.

Figures 3.5-3.9 show several theoretical sag detection responses. Detection thresholds,

set at ±2kV/s, are shown as bold lines that intersect with the RMS voltage deriv-ative signal. The point on wave effect on response time is evident when comparing

Figures 3.5 and 3.6. Here, the low magnitude sag of 20% shows a difference of 1.7ms

in triggering times between 0 and 90. Nevertheless, the sag detection algorithm

triggers in less than a quarter cycle - a significant improvement from using a basic

RMS calculation. Figures 3.7 and 3.8 show the theoretical sag detection algorithm

response when applied to higher magnitude sags of 80%. Detection times are still

below one quarter cycle, but are slightly increased for the 0 point on wave case.

RMS derivative threshold levels are critical in adjusting the detection response.

This is evident in Figure 3.9, where a sag of 89% does force the RMS voltage derivative

signal below its detection threshold. This case of a missed detection for shallow sags

may be alleviated by decreasing the detection thresholds, but at the cost of increased

sensitivity to transients and noise.

62

Figure 3.5 Simulated voltage sag (20%, 90) with RMS voltage, RMS voltage deriva-tive, and threshold signals.

Figure 3.6 Simulated voltage sag (20%, 0) with RMS voltage, RMS voltage derivative,and threshold signals.

63


Figure 3.8 Simulated voltage sag (80%, 0) with RMS voltage, RMS voltage derivative,and threshold signals.

64


Data Manipulation

Transferred Data

Data is transferred from the PLC to the Ridethrough PC in 16 byte (8 word)

blocks. Tables 3.3 and 3.4 detail the structure of these data blocks. A transfer occurs

during every Profibus network scan cycle, but the data updates to the PLC occur

asynchronously. The PLC program and peripheral scan rates contribute to the actual

update times for data sent to the Ridethrough PC, and the main Ridethrough PC

loop execution time contributes to the update rate for the data sent to the PLC.

Event Data

Event data is stored in a multidimensional array that increases its main index once

for each recorded event. Process signal data is stored in a series of subarrays that

are updated in a shift/discard technique similar to the array update method used to

monitor line voltages. In addition to these signal arrays, values of event length in

65

Table 3.3 Data transferred from PLC to ridethrough PC.

WORD DESCRIPTION0 Tension1 Status word2 Takeup tachometer3 Payoff tachometer4 Takeup output reference5 Payoff output reference6 Linespeed setpoint7 Tension setpoint

Table 3.4 Data transferred from ridethrough PC to PLC.

WORD DESCRIPTION0 Tension replacement value1 Control word2 Takeup tachometer replacement value3 Payoff tachometer replacement value4 Takeup output reference replacement value5 Payoff output reference replacement value6 Linespeed setpoint replacement value7 Tension setpoint replacement value

iterations and seconds is stored. Minimum RMS voltages and minimum RMS voltage

derivatives for each phase are also measured for storage in the main array. Lastly, a

Boolean array that indicates the phase combination of the sag is stored. Table 3.5

summarizes the values stored for each event. The storage trigger command occurs

1.110s after the beginning of a voltage sag, and captures 278ms of prefault data

along with 1.110s of fault/postfault data. These lengths were determined in practice

to be of sufficient length to capture all relevant process event data, while not affecting

program execution by placing excessive demands on system resources. This operation

only stores the data in memory; a file storage command is executed after the program

has terminated.

The goal of creating a large process history is for future reference in the event of an

incoming voltage sag. If measured reactions to sags are known, then the Ridethrough

PC may access the data and determine the expected response to an incoming sag. A

66

mitigation solution may then be applied to match the expected response. The final

size of the event data array is 216 elements. With a magnitude step size of 20%, this

stores an event for every allowable magnitude and phasing combination.

Table 3.5 Process data storage array elements.

ITEM DESCRIPTION0 Tension subarray (1667 elements)1 Status subarray (1667 elements)2 Takeup tachometer subarray (1667 elements)3 Payoff tachometer subarray (1667 elements)4 Takeup output subarray (1667 elements)5 Payoff output subarray (1667 elements)6 Linespeed setpoint subarray (1667 elements)7 Tension setpoint subarray (1667 elements)8 Length of event in iterations9 Minimum phase A RMS voltage (%)10 Minimum phase B RMS voltage (%)11 Minimum phase C RMS voltage (%)12 Minimum phase A RMS derivative voltage (V/s)13 Minimum phase B RMS derivative voltage (V/s)14 Minimum phase C RMS derivative voltage (V/s)15 Length of event in seconds16 Phasing combination Boolean array

Storage and Retrieval

Upon termination of the main program loop, event data may be recorded in a

text file. The file grows to a size of approximately 23MB for the desired 216 stored

events. Retrieval is performed at the beginning of program execution, and requires

a user input to access the data file. The text file data is parsed into memory in

single-event sized blocks. After completion of the read-in process, the original event

data array is rearranged into a 6 x 6 x 6 array, where each dimension of the new data

array represents an input voltage phase with varying magnitude (20% increments).

Process historical data is later retrieved through this array, where the indexing terms

correspond to the measured RMS phase voltage magnitudes.

67

Event Analysis and Association

Event characteristics are computed and stored using the procedure detailed in

Figure 3.10. Among these calculations are minimum sag magnitude, minimum RMS

voltage derivative values, sag length, and prefault signal averages. They are stored

after the event has ended, but are updated every program cycle.

When a voltage sag is detected, a magnitude and phasing analysis is performed.

This analysis takes one-quarter of a 60Hz cycle to complete. Once the sag magnitude

and phasing characteristics are determined, association with a recorded event and

response region identification may take place.

Sag Measurement

Sag magnitudes are measured with a peak detection routine. This is performed

with an analysis of the input power data arrays. Given that

VΦ(t) = Vm sin(ωt), (3.4)

andd

dtVΦ(t) = Vmω cos(ωt), (3.5)

we may be assured that the peak magnitude of the incoming voltage is present in either

a one-quarter cycle sample set of the raw voltage data or the rescaled derivatives

calculated with the same one-quarter cycle set. The aim is therefore to find the

maximum magnitude amongst two sets of data. The first is a 25 sample set of the

most recently acquired raw input voltage data,

VΦ1(k) = VΦ(k), k = 0...24, (3.6)

where k is the array index. The second set contains 25 scaled derivatives determined

by

VΦ2(k) =VΦ(k)− VΦ(k − 1)

ωTs, k = 1...25. (3.7)

where ω is the frequency in rad/s and Ts is the sampling period. The index k in

Equation 3.7 exceeds that of Equation 3.6 by one, which makes the overall required

68

START EVENTANALYZE

FOR EACH PHASE:dVRMS < MINIMUM

dVRMS?

YES

NO

MINIMUM dVRMSPHASE i = dVRMS

PHASE i

FOR EACH PHASE: VRMS < MINIMUM VRMS?

YES

NO

MINIMUM VRMSPHASE i = VRMS

PHASE i

WAS_ANY_SAG=FALSEAND IS_ANY_SAG=TRUE?

YES

NO

START ITERATION =PRESENT ITERATION

YES

NO

IS_ANY_SAG=TRUE?YES

NO

SAG LENGTH = 5T s x(ITERATION-START

ITERATION)

YES

NO

STORE BOOLEANSAG PHASINGINFORMATION

CALCULATEPREFAULT VALUES:TENSION, POTACH,

TUTACH,POCONTROL,TUCONTROL

ITERATION = STARTITERATION +1334

(1.1 SEC)?

YES

NO

STORE BUS DATAARRAYS, PHASING

DATA, SAG LENGTH,MIN VRMS, MIN

dVRMS, #ITERATIONSOF EVENT

RESET MIN VRMS=1,MIN dVRMS=0

END EVENTANALYZE

WAS_ANY_SAG=TRUE AND IS_ANY_SAG=FALSE?

END ITERATION =PRESENT ITERATION

IS_ANY_SAG=TRUE AND ITERATION=START

ITERATION+7(1/3 CYCLE)?

Figure 3.10 Flowchart of event analysis in ridethrough PC.

69

sample time one sample greater than a quarter cycle. The peak magnitude is then

determined with

Vm = max(|VΦ1| , |VΦ2|). (3.8)

Figures 3.11 and 3.12 show the theoretical response to the peak detect routines

given sags of 20% depth and 0 and 90 point on wave, respectively.

Event Association

After the phase voltage magnitudes have been determined, they are each divided

by their nominal value to determine the per unit voltage sag magnitude. These are

converted into integer array indices used to access the process history array. The

array indices are generated using a nearest integer function of the form

wΦ =Vp.u.ρv

, (3.9)

where wΦ is an array dimension index, Vp.u. the per unit voltage magnitude, and ρv

the per unit voltage step size of indexed sag magnitudes. With the Ridethrough PC

using 20% increments, the output range of this function is 0 ≤ wΦ ≤ 6.After indices are generated for each phase, they are used to access the 6 x 6 x 6 data

array to yield a process response recorded prior to the present voltage sag. Analysis

of this data may then be used to determine the mitigation method.

Identification of Response Region

In the previous chapter, three different voltage sag response regions were identified

for the textile tension control stand. They consist of an AC drive dropout region, a

faulted instrumentation region, and a no disruption region. The measured magnitude

and phasing combination is used to determine the response region of an incoming sag.

The user selects one of two possible identification methods.

Identification Using Unmitigated Process History

After retrieving unmitigated process response data that corresponds to an in-

progress voltage sag, the historical data is analyzed to determine the response region

in which it lies. For determining membership in the AC drive dropout region, the

70

Figure 3.11 Theoretical response of peak detect routine to 40%, 0 point on wavevoltage sag.

Figure 3.12 Theoretical response of peak detect routine to 40%, 90 point on wavevoltage sag.

71

recorded status word from the PLC is scanned for fault bits indicating drive dropout.

Membership in the faulted instrumentation region requires analysis of the recorded

fault/postfault signal data and comparison to the average prefault value. Upper

and lower tolerance thresholds are established, and the present sag is said to be

in the faulted instrumentation region if any of the critical signals fall outside these

thresholds. Upper and lower tolerance thresholds were set between ±5%-15% of

their recorded prefault values. Steady state operating conditions are assumed, and

the thresholds are left as user defined quantities. Identification of the no disruption

region is unnecessary because it requires no mitigation response. This determination

algorithm requires a fair amount of memory to store the history data block, but holds

the distinct advantage of not requiring mathematical models to identify response

regions. Figure 3.13 shows the flowchart for this method.

Identification Using Region Boundary Definitions

The region boundary definitions described in Equations 2.30, 2.31, and 2.32 may

be implemented in code to perform a less computationally expensive expected re-

sponse identification. This method does not require referencing historical process

data, but instead requires an estimation of device thresholds and region boundaries.

The flowchart for region boundary response identification is shown in Figure 3.14.

Again, specific identification of the no disruption region is unnecessary, as it does not

mandate a mitigation response.

72

USER SELECTED HISTORYRESPONSE ID AND

IS_ANY_SAG=TRUE ANDITERATION = START ITERATION +7

(1/3 CYCLE)?

YES

NO EXPECTED DRIVEDROPOUT = TRUE

NO

IS DRIVE DROPOUT PRESENT

IN RECORDEDEVENT?

YES

NO

NO EXPECTEDPROCESS

RESPONSE

EXPECTEDFAULTED

INSTRUMENTS =TRUE

CALCULATEHISTORY ARRAY

INDICES FOR EACHPHASE VOLTAGE

ACCESS PROCESSHISTORY ARRAY

WITH GENERATEDINDICES

DO ANY INSTRUMENT ORCONTROL SIGNALS FALL

OUTSIDE ACCEPTED DEVIATIONIN RECORDED EVENT?YES

START HISTORYRESPONSE ID

END HISTORYRESPONSE ID

Figure 3.13 Flowchart for identification of expected process voltage sag response usingprocess history.

73

STARTMEASUREMENTRESPONSE ID

USER SELECTED MEASUREMENTRESPONSE ID AND

IS_ANY_SAG=TRUE ANDITERATION = START ITERATION +7

(1/3 CYCLE)?

YES

NO

EXPECTED DRIVEDROPOUT = TRUE

MAG(A) + MAG(B) < 1.5p.u. ANDMAG(B) + MAG(C) < 1.5p.u. AND

MAG(C) + MAG(A) < 1.5p.u. ?

NO

YES

MAG(C) < 0.7p.u. ?

YES

NONO EXPECTEDPROCESS

RESPONSE

EXPECTEDFAULTED

INSTRUMENTS =TRUE

ENDMEASUREMENTRESPONSE ID

Figure 3.14 Flowchart for identification of expected process voltage sag response usingthreshold calculations.

74

Mitigation Routines

Once the expected response region of the incoming voltage sag is identified, then

one of four ridethrough routines may be executed. Two routines aim at mitigating

AC drive dropout faults and two address faulted instrumentation responses. When an

incoming sag falls into the no disruption region, then the process is left undisturbed

by the Ridethrough PC.

AC Drive Dropout Routines

Software Momentary Drive Coast

In the AC drive dropout region, the drives halt their output when their DC bus

magnitude crosses its trip threshold. As shown in the unmitigated response study of

the previous chapter, the rate of decent from the prefault magnitude to the trip level

is significantly less when the motor is disconnected than when the motor is energized.

The AC drive mitigation routines take advantage of this effect. Their aim is to switch

the drives to a coast condition if a sag is expected to cause a DC bus trip, then re-

energize them after the sag has completed. The flowchart of Figure 3.15 shows the

procedure for the software momentary drive coast.

When the voltage sag has ended, the average prefault control values are substi-

tuted for the output control signals. Using the prefault values is a means of gradually

returning the process to its normal state, instead of using feedback control which risks

an extreme process variable overshoot upon recovery. After 1.5 seconds has elapsed

on a postsag timer (a time sufficient enough to allow the process to return to normal

operation), the standard feedback control conditions are restored. During the coast

command, the PLC subroutine also calls for a PI control integrator hold. This allows

feedback control to restore to prefault conditions without any signal drift occurring

during the sag and process restart steps. The restarting algorithm flowchart is shown

in Figure 3.16.

Naturally, this approach is expected to cause a process disturbance. In the case

of the textile tension controller, the expected outcome is to experience a momentary

75

USER SELECTED SOFTWARE COASTAND ITERATION = END ITERATION + 2

AND HOLD COAST = TRUE ?

YES

NO

COAST COMMAND = FALSE,HOLD COAST = FALSE,

EXPECTED DRIVE DROPOUT = FALSE,RESTARTING = TRUE

USER SELECTED SOFTWARE COASTAND EXPECTED DRIVE DROPOUT =

TRUE AND IS_ANY_SAG = TRUE?

YES

NO

COAST COMMAND = TRUE,HOLD COAST = TRUE

START SOFTWARECOAST

END SOFTWARECOAST

Figure 3.15 Flowchart of software momentary drive coast mitigation routine.

loss of tension along with changes in linespeed. Figure 3.17 illustrates the simu-

lated process response to this mitigation method. The expected process disturbance,

however, is a significant improvement over the unmitigated response. Without the

ridethrough routine, the process shuts down, requires rethreading, and calls for a user

issued restart.

The software momentary drive coast routine requires that the coast command

reaches the drives before the DC bus crosses the trip threshold. Therefore, the

effectiveness of the software coast is highly dependent on the system latency. Im-

provements in execution times will allow more time for the DC bus to decay at the

de-energized rate before it crosses the trip threshold. The second AC drive dropout

mitigation routine explores improving coast command performance in this manner.

76

RESTARTING = TRUE ORHOLD CONTROLS = TRUEOR HOLD COAST = TRUE?

YES

NO

CONTROL SIGNALREPLACE BITS =

TRUE

CONTROL SIGNALREPLACE BITS =

FALSE

RESTARTING = TRUEAND ITERATION = END

ITERATION + 1800(1.5 SEC)?

YES

NO

RESTARTING =FALSE

EXPECTED FAULTEDINSTRUMENTS =TRUE AND

ITERATION = END ITERATION+400 (1/3 SEC)?

YES

NO

EXPECTED FAULTEDINSTRUMENTS =

FALSE

START RESTART /RESET

END RESTART /RESET

Figure 3.16 Flowchart of process restarting sequence following coast mitigation rou-tine.

77

Figure 3.17 Theoretical process response to AC drive coast mitigation routines.

Hardware Assisted Momentary Drive Coast

The hardware momentary drive coast routine bypasses the Profibus network and

PLC when performing an output halt to the payoff and takeup reel motors. This

is accomplished by using an output channel available on the data acquisition card.

When the halt command is given, the data acquisition card issues an output voltage

which drives a bank of relays that physically disconnect the motors from the drives.

In effect, this method performs the same operation as the software coast routine, but

does not require a coast and restart command to the PLC. The output control signal

replacement commands and the PI control integrators are controlled as they were in

the software routine. The algorithm flowchart is shown in Figure 3.18 and the add-on

circuit for this method is shown in Figure 3.19. This relay configuration was chosen

for its ability to be driven entirely by the data acquisition card’s analog output and

78

available power supply. The data acquisition card sends a 0VDC or 5VDC signal

from an analog output channel, which closes or opens an optical solid state relay.

Operation of the solid state relay in turn operates a bank of 5VDC electromechanical

relays that open and close the connection between the AC drives and the motors.

Excitation voltage for the electromechanical relays is provided by a 5VDC, 1A supply

from the data acquisition card. When the analog output is set to 0VDC, the drive-

motor connection is closed. For the 5VDC analog output, the connections are opened,

and the drive DC buses are prohibited from losing energy in the motors.

USER SELECTED HARDWARE COASTAND EXPECTED DRIVE DROPOUT =

TRUE AND IS_ANY_SAG = TRUE ?

YES

NO

HOLD COAST =TRUE

DISCONNECTMOTOR OUTPUT

RELAYS

USER SELECTED HARDWARE COASTAND ITERATION = END ITERATION +2

AND HOLD COAST = TRUE ?

YES

NO

HOLD COAST = FALSE,EXPECTED DRIVE DROPOUT = FALSE,

RESTARTING = TRUE

RECONNECT MOTOROUTPUT RELAYS

START HARDWARECOAST

END HARDWARECOAST

Figure 3.18 Flowchart of hardware assisted momentary drive coast mitigation routine.

79

10Ω

CRYDOMOPTICAL SOLIDSTATE RELAYG2-DB02 0113(NORMALLY

CLOSED)

NC

NC

NC

NC

DATAACQUISITION

CARD TERMINALBLOCK

ANALOG OUTPUT+5VDC/0VDC

ANALOG OUTPUTGND.

+5VDC SOURCE1A MAX

GND.

CONTROL CONTACTS

OMRONG6B-2114P

5VDCRELAY COILS

CR1

CR2

CR3

FROM PAYOFFMOTOR DRIVE

TO PAYOFFMOTOR

CR4

CR5

CR6

FROM TAKEUPMOTOR DRIVE

TO TAKEUPMOTOR

CR1

CR2

CR3

CR4

CR5

CR6

Figure 3.19 Add-on circuit for hardware assisted momentary drive coast mitigationroutine.

80

The benefit of this method is that the coast command may reach the drives sooner

than with the software routine, but at the cost of adding hardware. With a halt

occurring sooner, the DC bus is given extra time to decay at the de-energized rate

before the trip level is crossed. Improvements gained with this routine must therefore

be weighed against the effort required to install the add-on circuit, along with the

ability of the motor drives to permit an open-circuited motor output.

Faulted Instrument Mitigation Routines

Because responses in the faulted instrumentation region involve the propagation

of erroneous system signals, the mitigation routines in this section concentrate on

substitution of these signals with temporary values.

Control Signal Substitution

In this algorithm, if an incoming voltage sag is found to be in the faulted instru-

mentation region and the user selects the control signal substitution routine, then

the Ridethrough PC selects the payoff and takeup control words to be replaced with

their average prefault values. After a postsag delay of 1.5 seconds, the output signals

are switched back to originate from the PLC controllers. During the execution of

this routine, the PI integrators are held to allow for a smooth transition back to the

original control conditions. This method is similar to the open loop method proposed

in [68], but is only triggered for sags that fall into the faulted instrumentation region.

The instruction flowchart for this mitigation method is shown in Figure 3.20.

As with the AC drive dropout routines, original conditions are restored a fixed

amount of time after sag recovery. As shown in Figure 3.21, the expected response of

this routine keeps the physical quantities of linespeed and tension constant, assuming

steady state prefault operating conditions. This method does not take into account

the variations in ridethrough and recovery times that are present with different sen-

sors, for different sags, and for varying physical inputs. With increased specification

of individual sensor responses, signal replacement may be a more customized affair,

and allow the PI control loops to operate through a sag in this region.

81

USER SELECTED CONTROL SIGNALSUBSTITUTION METHOD AND EXPECTED

FAULTED INSTRUMENTATION = TRUE?

YES

NO

HOLD CONTROLS = TRUE

HOLD CONTROLS = FALSE

START CONTROLSIGNAL

SUBSTITUTION

END CONTROLSIGNAL

SUBSTITUTION

Figure 3.20 Flowchart of control signal substitution mitigation routine.

Selective Instrument Signal Substitution

In the selective instrument signal substitution routine, the average prefault values

of each sensor are determined and used along with the incoming sag magnitude to

access an array of ridethrough times for each instrument. The values in these arrays

are determined from previously stored sag testing results as recorded in Appendix E.

Array indices are generated using Equation 3.9 and the equation

wy =yavg − ymin

ρi, (3.10)

where wy is the physical input level array index for accessing the ridethrough time,

yavg is the average prefault value of the physical input, ymin is the minimum physical

value for the process, and ρi the step size in units of the physical input that instrument

i measures (m/s, N). With a physical input and sag magnitude known, the accessed

ridethrough time is used as a delay time for the instrument signal to be replaced with

82

Figure 3.21 Theoretical process response to control signal substitution and selectivesignal substitution routines.

the average prefault value. In cases where the instrument rode through an event,

an arbitrarily large delay time is stored in the array, which effectively prohibits the

instrument signal from being replaced. Maximum recovery times are also accessed for

each instrument, and after the sag is completed, the instrument signal is not routed

back to the control algorithm until the postsag timer has elapsed the instrument

maximum recovery time. A flowchart for this algorithm is shown in Figure 3.22.

The significant improvement this mitigation method makes over control signal

substitution is that the instrument signals are allowed to pass into the control algo-

rithm if their value is expected to be undisturbed. This allows for closed loop control

while the instruments are in the ridethrough segment of their sag response. For sags

of a shorter duration than the ridethrough times, the control algorithm is unaffected

and the inherent instrument ridethrough is used to its full capacity. Also eliminated

83

FOR EACH INSTRUMENT:IS_ANY_SAG = FALSE AND

TIME AFTER SAG > MAXIMUMRECOVERY TIME AND REPLACEINSTRUMENT SIGNAL = TRUE?

YES

NO

REPLACE INSTRUMENT SIGNAL =FALSE

YES

NO

CALCULATE EXPECTED RIDETHROUGHTIME ARRAY INDICES WITH VOLTAGE

SAG MAGNITUDE AND PREFAULTINSTRUMENT VALUES

USER SELECTED SELECTIVE SIGNAL SUBSTITUTION AND

EXPECTED FAULTEDINSTRUMENTS = TRUE AND

IS_ANY_SAG = TRUE ?

YES

NONO

REPLACE INSTRUMENT SIGNAL =TRUE

DETERMINE EXPECTEDRIDETHROUGH TIMESWITH ARRAY INDICES

YES

FOR EACH INSTRUMENT: ISPRESENT SAG LENGTH > EXPECTED

RIDETHROUGH TIME?

USER SELECTED SELECTIVE SIGNAL SUBSTITUTION AND

EXPECTED FAULTED INSTRUMENTS =TRUE AND ITERATION = START

ITERATION +7?

START SELECTIVESIGNAL SUBSTITUTION

END SELECTIVESIGNAL SUBSTITUTION

Figure 3.22 Flowchart of selective instrument signal substitution mitigation routine.

84

is the need to hold the PI control integrals for a smooth transition following recovery.

The theoretical response of this method as it is applied to the textile tension control

stand is the same as the theoretical control signal substitution response as shown in

Figure 3.21.

Summary

A software ridethrough coordination and mitigation program was developed to

interface with a three phase power supply and the Profibus network used on the

experimental textile tension control stand. Interfacing hardware consists of voltage

isolators and a National Instruments data acquisition card for the power supply, and

a Woodhead/SST Profibus interface card for the automation network. Additions to

the existing PLC ladder program were minimal, consisting of an add-on subroutine to

manage data transfer and a series of signal switching commands in the main ladder

program. Implementation and execution of the main ridethrough program occurs in

the LabView graphical programming platform.

Voltage sag detection is triggered by RMS derivative threshold crossings, with

a theoretical detect time of less than one-quarter cycle. Sags are measured in an

additional quarter cycle, and are either associated with historical process data corre-

sponding to the incoming sag or applied to logic functions that determine the expected

process response. The program monitors selected process data from the automation

network, and in the event of a sag stores signals and status data for future reference.

Voltage sag measurements are also taken and included in the event history.

Mitigation routines are executed based on preselected user input and the expected

disruption region of the incoming sag. Sags that are categorized in the AC drive

dropout region are mitigated by either a command driven coast routine or a contactor

driven coast routine. Command driven coasting is entirely software based, but is

potentially limited by data transfer rates. Contactor coasting can increase response

time by functioning outside the PLC and data network, but requires the addition of

motor contactors and switching relays.

85

Sags that are expected to cause instrumentation faults are addressed in one of

two ways. In the control signal substitution method, the motor drive control signals

are immediately replaced with their prefault values and the PI control integrators are

held. After the sag and a fixed waiting period have expired, the PI control outputs are

redirected to their original locations. In the selective substitution method, timers are

set to the ridethrough times for a given instrument. This timer setting is determined

by the present voltage sag magnitude and instrument physical input. Expiration of

a ridethrough timer triggers the instrument signal to be replaced with its average

prefault value. This is an improvement over the control signal substitution method

because it takes advantage of each instrument’s inherent ridethrough capabilities.

Detailed sag testing is, however, required to determine the appropriate delay times.

Instrument recovery times are also referenced to determine the optimal times for

switching the original signal back into the control loop.

Additional flowcharts for the Ridethrough PC programming is given in Appendix

I, and the LabView code is provided in Appendix J.

86

CHAPTER 4

EXPERIMENTAL RESULTS

The Ridethrough PC software as described in the previous chapter was applied to

the experimental textile tension control stand. This chapter provides an assessment

of software mitigation algorithm effectiveness, along with analysis of the interfacing,

sag detection, and event analysis/association subsystems.

Execution Times

The latency, or stimulus-to-response timing of the Ridethrough PC software may

be assessed in terms of component execution times. These components are detailed

in Figure 4.1, which shows the order of operational delays for each mitigation routine.

The measured times are listed in Table 4.1.

VOLTAGESAG IN PC DATA

TRANSFERSPC LOOP TIME

(MULTIPLE INSTANCES)

PROFIBUS DATA TRANSFER

PLC LOOP EXECUTION

HARDWARE ASSISTED MOMENTARY VFD COAST ROUTINE

SOFTWARE MOMENTARY VFD COAST ROUTINE

CONTROL SIGNAL SUBSTITUTION ROUTINE


DRIVE SHUTOFFDELAY

DONE

DATA ACQUISITION

CARD OUTPUT

RELAY OPENING TIME

DONE


PLC LOOP EXECUTION

DONE

PC DATATRANSFER

Figure 4.1 Order of operations for analysis of execution times.

87

Table 4.1 Individual operation execution times.

OPERATION EXECUTION TIMEPC data transfer 6µs

Ridethrough PC loop time 0.833msProfibus data transfer 5.912msPLC loop execution 8.7msDrive shutoff delay 25ms

Data acquisition card output time 204µsDrive output relay opening time 3.61ms

Timing Measurements

Measurement of the values in Table 4.1 were performed by the following methods:

1. PC data transfer. The delay time was determined by executing a program that

sends an incremental variable to the Profibus network. Examination of the

Profibus data stream for data value transitions yielded increment numbers and

their corresponding time stamps. For each data point, the increment number

and time increment was recorded. The average length per increment was de-

termined to be 6µs and is attributed to PC data transfer delays from the data

acquisition card or to the Profibus interface card.

2. Ridethrough PC loop time. This is defined by the sampling period for the

voltage signal inputs. Five data points are sampled for each loop increment at

a sampling frequency of 6kHz, yielding a loop time of 0.833ms.

3. Profibus data transfer. Loop cycle times were measured using Scope Profibus

software, and were determined to vary from an experimental low of 5.418ms to

a high of 7.722ms, with an average of 5.912ms.

4. PLC loop execution. Determination of PLC execution times were made by ac-

cessing the PLC through Siemens Step 7 programming and diagnostic software.

Scan cycle times varied between 6ms and 11ms, with an average of 8.7ms.

5. Drive output shutoff delay. The time required for the drive to respond to a

direct (non-network issued) shutoff command was determined to lie between

18ms and 31ms with an average response time of 25ms.

88

6. Data acquisition card output time. This time is determined by a executing loop

containing an analog output command only. The average time is calculated

using the number of loop iterations over a fixed period of time, which yielded a

value of 210µs. The PC data transfer delay is subtracted from this value, which

results in a delay time of approximately 204µs

7. Drive output relay opening time. Experimentally this delay time varied from

3.48ms to 3.90ms with an average of 3.61ms.

Combined Routine Execution

The collective propagation and execution delays itemized in Table 4.1 may be

used to estimate the execution times for the software mitigation algorithms. These

estimated times are listed in Table 4.2. For each value, 12 Ridethrough PC program

cycles are included to account for sag detect and measurement times. Profibus net-

work delays and PLC loop execution times are each multiplied by 1.5 times their

measured average, which accounts for the full cycle where a response is computed

plus an average of one-half cycle when the stimulus arrives but is not recognized until

the beginning of the next full computation cycle. The selective signal substitution

execution times are not listed due to the variation caused by built-in time delays,

though the time-invariant components are identical to the control signal substitution

method.

Comparison of the estimated execution times in Table 4.2 with their experimental

counterparts in Table 4.3 reveals several differences. In the software momentary drive

coast algorithm the estimation falls above the experimental range, but the hardware

assisted drive coast algorithm is predicted to be faster than the experimental val-

ues. This may be attributed to unmeasured variation in the Ridethrough PC data

transfer times resulting from increased demand of system resources that the complete

mitigation program has over the experimental timing measurement program. Such

variations serve as an indicator that a computationally inexpensive software platform

is a key component of a software-based mitigation system. When comparing the

coast mitigation algorithms to one another, the improvement in execution time that

89

the hardware assisted method offers over the software-only method emphasizes the

dependence that algorithm effectiveness has on network communication throughput

and execution timing. The measurements made for the control signal substitution

method are made without the Profibus data transfer and PLC loop execution compo-

nents, but include the data acquisition card output component. For this algorithm,

if the experimental value in Table 4.3 is compensated with the average measured

component values to account for the measurement method, the execution time range

lies between 33.4ms and 35.4ms, which is a closer match to the predicted value, but

remains slightly longer.

Table 4.2 Mitigation routine predicted execution times.

MITIGATION ROUTINE APPROX. EXECUTION TIMESoftware momentary drive coast 65.9ms

Hardware assisted momentary drive coast 13.9msControl signal substitution 32.0ms

Table 4.3 Mitigation routine experimental execution times.

MITIGATION ROUTINE ACTUAL EXECUTION TIMESoftware momentary drive coast 56ms ≤ t ≤ 64ms

Hardware assisted momentary drive coast 20ms ≤ t ≤ 45msControl signal substitution 19ms ≤ t ≤ 21ms

90

Performance of Sag Detection, Sag Measurement, and Event Association

Sag Detection

Experimental performance of the voltage sag detection method is shown in Figures

4.2-4.5 and summarized in Table 4.4. In several cases, the actual detect times are

one to two execution cycles of the Ridethrough PC program (multiples of 0.833ms)

longer than is predicted by MATLAB simulations (Figures 3.5-3.8). When comparing

the experimental and simulated cases, the RMS voltage derivative deviates from its

prefault value more rapidly in the simulated case. This result is caused by slight

differences in nominal voltage magnitude and distortions from a true sine wave in

the experimentally captured input waveform. The distortion’s effects are evident in

the experimental prefault RMS voltage derivative waveforms, where a slight ripple

can be observed. Differences between actual and predicted response times are also

caused by the time at which the sag is initiated relative to the execution cycles of

the Ridethrough PC. In the experimental cases, the sag is initiated at a random time

relative to the Ridethrough PC execution. This asynchronous behavior leads to RMS

voltage derivative threshold crossings that are detected at slightly different times,

which can lead to a single Ridethrough PC execution cycle difference of 0.833ms.

Cases of 89% magnitude, 90 point in wave (Figure 4.6) and 89% magnitude,

0 point in wave voltage sags were also observed. In both examples, no detection

signal is generated, which matches the simulated prediction (Figure 3.9 shows the

89%, 0 simulation). This result can be alleviated using RMS voltage derivative

detection thresholds lower than the settings of ±2kV/s. This change must take intoconsideration the balance between fast sag measurement and low susceptibility to line

noise and voltage transients. If these tolerance bands were narrowed, the resulting

increase in sag detection performance would come at the cost of increased sensitivity

and false sag detection signals.

Experimentally, the voltage sag detection method proves to be a satisfactory solu-

tion for the system to which it is applied. It is computationally inexpensive and will

reliably detect events delivered by the voltage sag generator that supplies the tex-

tile tension control process. Detections occur within 5.0ms of voltage sag inception,

91

which provides adequate time for measurement and expected response identification

to occur before the process is disrupted.

Table 4.4 Actual and theoretical voltage sag detection times.

SAG MAG. / P.O.W. THEORETICAL TIMES ACTUAL TIMES20% / 0 2.50ms on, 1.67ms off 3.33ms on, 2.50ms off20% / 90 0.83ms on, 0.83ms off 0.83ms on, 0.83ms off80% / 0 3.33ms on, 3.33ms off 5.00ms on, 4.17ms off80% / 90 0.83ms on, 0.83ms off 2.50ms on, 1.67ms off89% / 0 No detections No detections89% / 90 No detections No detections

Figure 4.2 Experimental detection response to 20%, 90 point on wave voltage sag.

92



93



94

Sag Measurement

Graphs of the experimental voltage sag measurement output are shown in Figures

4.7 and 4.8. In the example for a 40%, 0 point in wave voltage sag, the resulting

magnitude measurement possesses the expected time delay of one-quarter cycle be-

fore the measurement settles to a reliable value. The experimental case verifies the

simulated case (Figure 3.11), where the sole difference between experimental and the-

oretical cases is a slight deviation in the experimental sag magnitude measurements

at the points of inception and recovery. Despite this difference, the measurement is

reliable for the application of utilizing the sag magnitude calculation after allowing

the one-quarter cycle delay to elapse.

For the 40%, 90 point in wave case shown experimentally in Figure 4.8 and

theoretically in Figure 3.12, both cases show the resulting magnitude measurement

spike caused by abrupt voltage transitions applied to the calculation of Equations

3.7 and 3.8. This underscores the criticality of allowing a one-quarter cycle delay to

elapse before applying the measurement towards identifying the voltage sag’s likely

process effects. Naturally, line noise present in the system will contribute towards

errors in measurement of the voltage peak using this method. This issue may be

alleviated with hardware or software filters, the application of which should involve

specific consideration of the filter performance versus the delays that they introduce.

Nevertheless, this measurement method proves to be a simple and effective means of

determining a voltage sag’s magnitude shortly after it is detected.

95

Figure 4.7 Experimental magnitude measurement response to 40%, 0 point on wavevoltage sag.

Figure 4.8 Experimental magnitude measurement response to 40%, 90 point on wavevoltage sag.

96

Event Association

The event association methods are executed seven Ridethrough PC cycles after a

detect signal is given, which allows sufficient time for the peak voltage measurement

to reach its final value. The method of identifying the expected response region is

preselected by user input. For the process history identification method, the history

is read into memory before the program begins executing, which avoids the delays

associated with opening and accessing a data file while the sag is in progress.

The method for process response identification using a recorded process history

functions properly; for the magnitude and phasing combinations delivered by the

voltage sag generator (in 20% increments), the detection and measurement routines’

application towards identifying a recorded event using Equation 3.9 reliably calculate

array indices to reference a detailed process voltage sag response history. Similarly,

identification of the expected process response using Boolean response region esti-

mations as described in Equations 2.31 and 2.32 performs properly and matches the

expected responses predicted using the process history. Under both of these meth-

ods, expected response determination occurs in less than one cycle of 60Hz incoming

voltage, which is significantly less than the delay time between voltage sag inception

and any unmitigated process response.

With a complete process response history using voltage sag magnitude increments

of 20%, the history file size is 23MB. For the identification of an expected drive

dropout, much of this data is unused. Only the state of the takeup and payoff drives

is scanned to determine membership in the drive dropout region and as such may be

reduced if only an expected drive dropout is to be determined. However, for identifi-

cation of expected faulted instrumentation, the recorded process instrumentation and

control signals are also scanned to determine out of tolerance conditions and therefore

require more detailed process history recording. In contrast, the memory requirement

of a process history is eliminated when using Boolean response boundary estimation,

but requires a clear definition of these boundaries to be effective.

97

Mitigation Algorithms

Each software mitigation algorithm was implemented and applied to the textile

tension control stand. This section describes the performance of each routine under

various test conditions.

Momentary Drive Coast Routines

The drive dropout routines aim at mitigating the dropout response as shown in

Figure 4.9. This is accomplished by entering a coast state in order to prevent or delay

the drive DC bus from crossing the trip level. Tests performed for these routines used

equal magnitude voltage sags on all three supply phases.

Software Momentary Drive Coast

Successful execution of the software momentary drive coast algorithm is shown

in Figure 4.10. Process setpoints are maintained at 1.0m/s line speed and 30N web

tension. In this case, the drive DC bus voltages are held above the trip level of

200V by temporarily stopping the drive output starting approximately 56ms after

sag inception. The DC bus decays at a slower rate, which provides additional time

for sag recovery to occur. After the sag has ended, both drives resume their output,

but are set to their prefault average setpoints (Figure 4.11). Product spillage and

manual restart are avoided, and the process line speed and tension return to their

prefault values and ultimately transition back to standard control at t = 1.95s. The

theoretical model predicts this behavior generally, but suggests a faster recovery than

in the experimental case. This difference is caused by the slack condition present in

the mechanical system that is not included in the theoretical model.

Figures 4.12 and 4.13 show the behavior of the software momentary drive coast

algorithm when recovery is not successful. In this case, the coast commands are

executed later than in the successful case because of variation in voltage sag detec-

tion times and command execution latency. The DC buses do not remain above the

dropout level of 200VDC for the entire duration of the interruption. Because of the

drive undervoltage faults, the restart and restore commands are executed with no ef-

fect. The unsuccessful execution of this algorithm ultimately mirrors the unmitigated

98

response and reveals the limitations of this method for sags of longer duration. This

scenario underscores the need for fast execution, but reveals that the unsuccessful

case is the inevitable result as the voltage sag duration approaches infinity, regardless

of the mitigating effort applied.

Altogether this mitigation strategy meets expectations — the process sees a dis-

turbance similar to a shutdown, but the shutdown is one that is intentionally caused

so that it can be brought back online in a controlled fashion. Process variables are

not continuously maintained, yet are restored without manual operator intervention.

Success of the software momentary drive coast routine requires that the coast com-

mand reaches the drives before the DC bus crosses the trip threshold and that the

voltage sag or interruption is of a reasonably brief duration such that the DC bus

voltages do not cross over their trip thresholds, even while decaying at a slower rate.

Despite its experimental success, the response times of this mitigation routine do

leave room for improvement. The effectiveness of this method is highly dependent on

Profibus network latency and PLC execution time. Faster coast executions will cause

DC buses to begin the de-energized rate of decay at a higher level and therefore allow

more time before the trip thresholds are crossed. If this method were implemented

such that its performance were less reliant on the propagation of coast commands over

the Profibus network from Ridethrough PC to PLC and then back to the VFDs, then

the momentary coasting routine would be tolerant to voltage sags of longer duration

and hence reduce the occurrences of unsuccessful execution. This reliance on network

and PLC timing gives rise to the hardware assisted momentary drive coast mitigation

routine.

99

Figure 4.9 Experimental and simulated unmitigated process response resulting indrive dropout (1.0m/s, 30N process setpoints; 0%, 450ms, 3φ voltage interruption).

100

Figure 4.10 Experimental and simulated process response using software momentarydrive coast algorithm (1.0m/s, 30N process setpoints; 0%, 450ms, 3φ voltage inter-ruption).

101

Figure 4.11 Experimental and simulated DC bus magnitude, motor output, andcontrol signal replace command during software momentary drive coast algorithm(1.0m/s, 30N process setpoints; 0%, 450ms, 3φ voltage interruption).

102

Figure 4.12 Experimental and simulated process response demonstrating unsuccess-ful process recovery during application of software momentary drive coast algorithm(1.0m/s, 30N process setpoints; 0%, 450ms, 3φ voltage interruption).

103

Figure 4.13 Experimental and simulated DC bus magnitude, motor output, and con-trol signal replace commands during unsuccessful software momentary drive coastalgorithm (1.0m/s, 30N process setpoints; 0%, 450ms, 3φ voltage interruption).

104

Hardware Assisted Momentary Drive Coast

Figure 4.14 shows an experimental case of the hardware-assisted drive coast rou-

tine for a three-phase voltage interruption of 0% magnitude applied for 450ms du-

ration. Both linespeed and tension responses are similar in form to the software

coast routine. The slack condition again causes a difference between the experimen-

tal and simulated recovery responses. For the 450ms sag in Figure 4.14, the difference

between the beginning of experimental and theoretical tension recovery is approxi-

mately 650ms, whereas the difference for the shorter sag duration case of Figure 4.15

is approximately 250ms. This indicates that for greater durations of zero tension

and coasting linespeed, more slack is created between the payoff and takeup reels.

Because of this, a delayed recovery occurs because the textile slack must be taken up

before tension is restored.

This difference reveals noteworthy characteristics of the process dynamics while

recovering from the coasting state. Substitution of prefault average control signals

lead to an overshoot in measured line speed while the slack is taken up, followed

by a return to the desired line speed while the web tension rises to its desired value

and gradually applies load to the motors. When slack is not accounted for, the

theoretical test runs will predict an overshoot in tension instead of line speed because

the mechanical coupling between rollers and reels is considered to be always present.

For the hardware momentary drive coast algorithm, the coasting state response

times were found to lie between 20ms and 45ms. The execution of the drive halt

operation is improved over the software momentary coast cases, where the successful

case response time is 56ms and the unsuccessful case response time is 64ms. The DC

bus voltage magnitude is held higher than in the software case, thereby allowing more

time to elapse before the undervoltage trip level crossover. Because of the difference

in DC bus decay rates between the energized motor and de-energized motor, the

improvement in response time pays off greater increases in allowable sag duration.

This improvement may be quantified as

∆t2 =A1A2− 1 ∆t1; |A1| > |A2| (4.1)

105

where A1 and A2 are linearly approximated slopes of the DC bus decay rates for

the energized and de-energized motor, and ∆t1 and ∆t2 are time delay intervals as

shown in Figure 4.16. When the software coast algorithm is used, the improvement in

voltage sag duration susceptibility increases by ∆t2. Additional improvements gained

using the hardware coast algorithm over the software coast algorithm may also be

quantified with Equation 4.1, where the value of ∆t1 instead represents the additional

improvement in response time over the software coast method.

Another remarkable outcome is that of the unnecessary operation case, as shown

in Figure 4.17. In this scenario, the momentary drive dropout mitigation routines

(both software and hardware-assisted) cause a brief coast condition, but the incoming

voltage sag recovers before the DC buses would cross the undervoltage trip level if

the routine were not applied. The observed process disturbance is therefore caused

entirely by the mitigation algorithm and creates a less desirable result than a process

voltage sag ride through. Though this condition will occur for a small set of possible

voltage sag durations, potential fixes such as building delay times into the mitigation

algorithms to minimize occurrences of the unnecessary case would be done at the

expense of any gains in sag duration tolerance that the mitigation routines seek to

achieve.

106

Figure 4.14 Experimental and simulated process response using hardware assistedmomentary drive coast algorithm (1.0m/s, 30N process setpoints; 0%, 450ms, 3φvoltage interruption).

107

Figure 4.15 Experimental and simulated process response using hardware assistedmomentary drive coast algorithm (1.0m/s, 30N process setpoints; 20%, 200ms, 3φvoltage sag).

108

Figure 4.16 Comparison of DC bus decay rates during momentary drive coastingalgorithm executions.

109

Figure 4.17 Experimental and simulated process response unnecessarily using hard-ware assisted momentary drive coast algorithm (1.0m/s, 30N process setpoints; 20%,200ms, 3φ voltage sag).

110

Faulted Instrumentation Routines

The faulted instrumentation routines mitigate the process deviation response by

interrupting and substituting feedback and control signals in the PLC controller. The

characteristic unmitigated faulted instrumentation response is shown in Figure 4.18,

where the applied fault is a single phase voltage interruption on Phase C of the power

system.

Control Signal Substitution

The control signal substitution routine replaces the drive setpoint control signals

with their prefault averages when faulted instrumentation is expected. Integral com-

ponents of the PI controllers are held to avoid accumulation of controller error. The

experimental and theoretical responses of the textile tension control stand using this

algorithm are shown in Figure 4.19. In this case, the AC motor drives do not drop

offline, but instead the tension sensor and tachometer produce faulted, erratic signals.

In response, the control signal replace bits are set (Figure 4.20) 19ms into the event

and for a delay of 333ms following the sag. The transition back to standard control

occurs at t=805ms. Algorithm execution results in the maintenance of both linespeed

and textile tension for the voltage sag event and hence a process ride through response.

Substituted signal transitions occur without any resulting disturbance in the process

outputs. However, during the open loop/signal substitution period, it is important

that no mechanical disturbance requiring closed loop control occurs in the process.

If this condition is met, the control signal substitution algorithm demonstrates an

effective means of improving process ridethrough in cases of faulted instrumentation.

As Figure 4.21 illustrates with a 50ms, 1φ interruption, there exists an unneces-

sary case response when applying this method. In this scenario, all instrumentation

rides through the event because the incoming voltage sag is not of sufficient dura-

tion to exhaust the instruments’ ridethrough for the applied voltage sag depth and

process physical conditions. No misoperation occurs in either the tachometer or the

tension cell, yet detection and measurement of the sag triggers the algorithms’s ex-

ecution. Signal substitution occurs between t = 21ms and t = 395ms. Unlike the

111

unnecessary case seen with momentary drive coasting routines, no resulting process

disturbance occurs. While occurrences of such a scenario are limited to voltage sags

of brief duration, they demonstrate the absence of customization with regard to in-

dividual instrument voltage sag characteristics. Therefore, while the functionality of

this routine is sound, room for improvement exists to the extent that variations in

instrument voltage sag responses may be accounted for when determining the need

to substitute process signals.

112

Figure 4.18 Experimental and simulated unmitigated process response resulting infaulted instrumentation (1.0m/s, 30N process setpoints; 0%, 450ms, 1φ voltage in-terruption).

113

Figure 4.19 Experimental and simulated process response using control signal substi-tution algorithm (1.0m/s, 30N process setpoints; 0%, 450ms, 1φ voltage interrup-tion).

114

Figure 4.20 Experimental and simulated faulted instrument signals and control replacebits (1.0m/s, 30N process setpoints; 0%, 450ms, 1φ voltage interruption).

115

Figure 4.21 Experimental and simulated process response unnecessarily using controlsignal substitution algorithm (1.0m/s, 30N process setpoints; 0%, 50ms, 1φ voltageinterruption).

116

Selective Signal Substitution

The selective signal substitution algorithm is designed to minimize the replacement

of instrument feedback signals in cases where fault and process conditions will not

cause an instrument error. The input variables of physical instrument inputs and

voltage sag magnitude are used to index arrays where the appropriate ridethrough

time is found and used as a signal switching delay.

The textile tension controller unmitigated case for 1.0m/s, 10N process inputs was

shown in Figure 4.18, where the faulted tension and tachometer inputs are shown to

produce a process disturbance for both linespeed and tension. Using the selective

signal substitution algorithm, the mitigated response for these conditions is shown in

Figure 4.22. Here both the experimental and theoretical responses indicate faulted

instrumentation signals, yet no appreciable linespeed or tension disturbances result.

The signal replace intervals are shown in Figure 4.23 and illustrate different delay

times for the initial signal substitutions as well as different delay times for restoring

the original signal feeds.

A test case for different physical input levels of 2.5m/s linespeed and 50N tension

is shown in Figures 4.24 through 4.26. The unmitigated response of Figure 4.24 shows

a significant disturbance for a single phase voltage interruption of duration 450ms.

Figures 4.25 and 4.26 show the selective signal substitution method functioning as

predicted, with no resulting process disturbance and signal transition times that are

customized to the input disturbance and process prefault characteristics.

117

Figure 4.22 Experimental and simulated process response using selective signal sub-stitution algorithm (1.0m/s, 30N process setpoints; 0%, 450ms, 1φ voltage interrup-tion).

118

Figure 4.23 Experimental and simulated faulted instrument signals and replace bitsfor selective signal substitution algorithm (1.0m/s, 30N process setpoints; 0%, 450ms,1φ voltage interruption).

119

Figure 4.24 Experimental and simulated unmitigated process response resulting infaulted instrumentation (2.5m/s, 50N process setpoints; 0%, 450ms, 1φ voltage in-terruption).

120

Figure 4.25 Experimental and simulated process response using selective signal sub-stitution algorithm (2.5m/s, 50N process setpoints; 0%, 450ms, 1φ voltage interrup-tion).

121

Figure 4.26 Experimental and simulated faulted instrument signals and replace bitsfor selective signal substitution algorithm (2.5m/s, 50N process setpoints; 0%, 450ms,1φ voltage interruption).

122

A case where only a single instrument is substituted is shown in Figures 4.27

through 4.29. In the unmitigated case, the tension cell measures a web tension of

10N and throughout the event provides a reasonable measurement that does not con-

tribute to the process disturbances. The tachometer, however, does deliver a faulted

signal and hence calls for replacement during the 40%, 450ms, 1φ supply voltage sag.

Upon applying the selective signal substitution algorithm, the linespeed and tension

disturbances are eliminated, as shown in Figure 4.28. The instrument signals and

switching signals shown in Figure 4.29 indicate that this particular execution of the

selective signal substitution algorithm did not include substitution of the tension cell

signal, thereby allowing it to function in the control loop in its normal capacity. A

noteworthy aspect of this test case is that at first inspection of the unmitigated re-

sponse it might seem that the tension cell misoperates and contributes to the process

disturbance. However, the instrument ridethrough time arrays indicate a ridethrough

condition for the given fault scenario and therefore prove themselves a more reliable

means of determining the appropriate signals to replace.

Experimentation indicates that the selective signal substitution method functions

adequately and is an improvement over the control signal substitution method because

the process controller is not bypassed unless absolutely necessary. The inherent ride-

through that exists in each device is used to its full advantage, which optimizes the

amount of time the PLC controller may be left unassisted before intervention. The

three salient test cases are compared in Figure 4.30, which illustrates the variation in

signal substitution intervals for each process response scenario.

Trivial cases of selective signal substitution algorithm execution also exist. When

the incoming voltage sag is determined to cause a possible faulted instrumentation

response, but the voltage sag duration and process variable conditions do not lead to

any signal replacement, the routine nonetheless executes, begins timing for replace-

ment commands, but never sends them. In these trivial case executions, closed loop

control is maintained throughout the duration of the incoming voltage sag event. This

ultimately leads to no mitigating action, no process disturbance, and maintenance of

all instrument signal feedback paths into the process controller.

123

Figure 4.27 Experimental and simulated unmitigated process response resulting infaulted instrumentation (1.0m/s, 10N process setpoints; 40%, 450ms, 1φ voltagesag).

124

Figure 4.28 Experimental and simulated process response using selective signal sub-stitution algorithm (1.0m/s, 10N process setpoints; 40%, 450ms, 1φ voltage sag).

125

Figure 4.29 Experimental and simulated faulted instrument signals and replace bits forselective signal substitution algorithm (1.0m/s, 10N process setpoints; 40%, 450ms,1φ voltage sag).

126

Figure 4.30 Comparison of process responses and signal substitution intervals forselective signal substitution algorithm executions.

127

Summary

An analysis of the Ridethrough PC operation was performed as it applies to the

test bed textile tension control system. Execution times were evaluated and found to

be strongly dependent on network delay and process controller cycle times. Detection

methods were found to trigger slightly longer than predicted, yet still prove effective

when used for the Ridethrough PC application. Measurement methods were found

to function as predicted, yet show sensitivity to waveform distortions and voltage

transients.

The software and hardware coast mitigation routines perform properly in their

mitigation of voltage sags and interruptions that would cause ACmotor drive dropout.

The software momentary drive coast method shows a strong dependence on network

latency and processing delays. Execution times for the hardware coast mitigation

method showed a significant improvement over the software coast method, but at the

cost of minor hardware additions. Both coast methods halt the motor outputs in

order to prevent the decaying DC bus from crossing its trip threshold, and cause a

brief loss of tension and linespeed decay. Recovery occurs in a gradual fashion and

eliminates extreme overcompensation caused by the accumulation of integral control

error. In cases of extended sag duration, a trip threshold crossover will occur even if

the mitigation routine is applied. Therefore, the effectiveness of the coast routines’

mitigating efforts may be quantified by the additional time the algorithms provide

before a drive DC bus undervoltage fault occurs. Unnecessary uses of the drive coast

routines were shown to exist, but their presence is restricted to sags that are of shorter

duration than the unmitigated dropout time.

When applied to the textile tension control system, the signal substitution rou-

tines were shown to improve the faulted instrumentation process response and agree

with their predicted outcomes. In tests of the control signal substitution method,

faulted instrument signals were blocked from propagating to the process drives at the

output stage. Arbitrarily substituting the output controls proves effective, but reveals

an unnecessary case of operation at times when a voltage sag recovers before instru-

ment ridethrough times elapse, or when physical input and voltage sag magnitude

128

conditions do not produce instrument faults. The unnecessary case is minimized by

the case of the selective signal substitution algorithm, which shows variation in the

use of and timing of switching prefault average instrument outputs into the process

controllers. This method is an improvement over control signal substitution for all

expected faulted instrumentation cases, as it allows the instrumentation to remain

in service when its misoperation is not anticipated. Trivial cases of selective signal

substitution were found to occur, but do not result in any corrective action.

129

CHAPTER 5

CONCLUSIONS AND FUTURE WORK

Investigation Conclusions

This dissertation presents novel software based means of mitigating the effects of

voltage sags on network connected industrial processes. Underlying design criteria

are to avoid the use of conventional mitigation hardware and to minimize disturbing

the existing system to which the solution is applied.

A dedicated textile tension control process is used as a test bed for applying the

software algorithms. Voltage sag testing on this process indicated that one of three

distinct responses can occur: AC drive dropout leading to process shutdown, faulted

instrumentation leading to process variable deviation, or no observed response (ride

through). Expressions for mathematically predicting the process response are derived,

where inputs are voltage sag magnitudes for three input power phases and outputs

are Boolean variables indicating an expected response.

The mitigation algorithms are applied using an add-on PC, or ‘Ridethrough

PC’, that interfaces with the supply voltages and the Profibus automation network.

Through monitoring and recording of network traffic during a voltage sag, a de-

tailed process history is established as an additional means of identifying the expected

process response. This history may be used independently of the mathematical ex-

pected process identification.

Using the Ridethrough PC, voltage sags are detected using an analysis of RMS

voltage derivatives and their magnitudes are measured using modified peak detection.

The expected process response is determined either through analysis of the process

history or Boolean calculation of the expected response.

Four voltage sag mitigation algorithms are proposed. A coast algorithm that is

exclusively software controlled addresses the AC drive dropout response. This method

preemptively forces the motor drives to coast in order to delay an undervoltage trip

130

condition. An anticipated temporary loss of tension and drop in linespeed results,

but the process recovers after the sag has ended, thus avoiding significant downtime,

product spillage, and user restart/reset of the process.

The software coast algorithm provides improved ridethrough for the process, but is

nevertheless susceptible to incoming sags of longer duration. Decreasing susceptibility

to longer duration voltage sags is achieved with the hardware assisted momentary

coast algorithm, which forces motor drive output contactors to an open circuit state

as a means of forcing the process to coast. Additional ridethrough beyond the software

coast algorithm is provided by an increased response time that stems from eliminating

the automation network and PLC delay times.

The faulted instrumentation process response is mitigated by one of two algo-

rithms. In the control signal substitution method, the output controls are held to

their prefault average values for the duration of the sag plus a postfault delay time.

During this time, the PI controller integrals are held to prevent an unnecessary ac-

cumulation of integral error before the signal routing is restored. This algorithm

performs well in maintaining process conditions during expected faulted instrumen-

tation voltage sags, but often performs unnecessary executions when the combination

of voltage sag magnitude, sag duration, and device physical input conditions will not

cause an instrument fault.

Improvements in faulted instrumentation methods are made in the selective sig-

nal substitution mitigation algorithm, which accesses stored ridethrough times as a

function of voltage sag magnitude and instrument physical input for use as signal sub-

stitution switching delays. Maximum recovery times are also used as times to switch

the signals back to their original routing. Experimentation with this algorithm proved

successful; signal substitutions are made after an instrument’s ridethrough time has

expired and are not made when instrument disturbances are not anticipated by the

event conditions. This algorithm effectively mitigates cases of expected faulted in-

strumentation up to the maximum test sag length of 450ms, maximizes the amount

of time an instrument feedback signal may remain in a feedback path, and reduces

instances of unnecessary signal substitution.

131

Development of the Ridethrough PC and the software-based voltage sag mitiga-

tion algorithms combined with experimental verification applied to a test bed textile

tension control workstation yields the following conclusions:

1. Use of software algorithms that execute in the presence of voltage sags can be

an effective means of mitigating process disturbances caused by voltage sags.

2. A recorded process response history or a mathematical response estimation may

be used to determine the appropriate course of action for software based voltage

sag mitigation.

3. AC drive dropout failures may be mitigated by application of software issued

coast commands. This solution increases the allowable time that an AC drive

dropout sag may last. Maintenance of process parameters during the coast state

is dependent on the specific process mechanics, but the ability to automatically

restore these parameters is enhanced by application of the coast routines.

4. Use of external contactors to force the coast state, as in the hardware assisted

momentary drive coast algorithm, improves the sag duration tolerance of the

process by eliminating the coast command delays of the automation network,

PLC program, and AC drive. This would not be the case if the latency of the

command path was less than the operating time of the external contactor.

5. Cases of unnecessary drive coast algorithm execution will occur when incoming

voltage sags are of a magnitude that will ultimately cause a dropout condition,

but are of a sufficiently brief duration such that the undervoltage fault would

not result. These cases produce process disturbances rather than mitigate them.

6. Arbitrary substitution of control signals during an incoming voltage sag that is

anticipated to cause a faulted instrumentation condition may lead to unneces-

sary substitution of control signals and excessive instances of forced open loop

control.

132

7. Implementation of predetermined instrument ridethrough and recovery times in

the decision making process for signal substitution can decrease the instances

where feedback signals are unnecessarily ignored and help preserve closed loop

process control in the presence of supply voltage sags.

8. Ridethrough improvements made with these strategies are constrained by com-

munication network latency and controller execution delays. These delays may

be reduced if execution bottleneck paths are bypassed.

Research Contribution

This investigation has focused on several topics that to date remain unexplored:

1. Existing software ridethrough solutions use only the presence of a voltage sag

as a means of triggering alternate control. The methods herein use measurable

sag characteristics as criteria for determining a required action.

2. A process history is utilized for the purposes of voltage sag mitigation. Process

variables and voltage sag event data are recorded and stored together. The

unmitigated process response is analyzed during an incoming event to determine

an expected response.

3. Process voltage sag response boundaries are expressed mathematically. These

expressions are then evaluated during in-progress voltage sags to determine the

expected process response for mitigation purposes.

4. Measured variations in instrumentation responses are used in determining the

need for momentary open loop control. Sag magnitude and physical input

levels serve as inputs in determining the allowable delay time before substituting

feedback signals. This approach utilizes the available ridethrough in each device

and minimizes the time open loop control is used.

133

Future Research Opportunities

System Implementation

Future work in this area should initially aim at improvements in implementation.

Programming the Ridethrough PC in a high-level programming language can reduce

the computational expense of the graphical LabView software. Faster and higher

resolution data acquisition hardware may be used to improve voltage sag detection

and sag measurement. To the extent that process controllers will permit, the software

may reside primarily in a process PLC to reduce the dependency on supervisory

systems.

More advancements may be found in the application of fuzzy logic or neural net-

work control that functions in a secondary or backup controller capacity in the pres-

ence of a supply voltage sag. It is likely that these solutions are more feasible in a

process that does not include a PLC and exclusively uses PC control. In this modified

architecture, the additional Ridethrough PC may be unnecessary, making application

of software mitigation algorithms less reliant on fieldbus network traffic.

Application of Commercially Available Equipment

Complexity of the Ridethrough PC can be reduced by using available measure-

ments from commercial metering or protective relaying equipment. This modification

would potentially eliminate the requirement for data aquisition hardware in the Ride-

through PC and the task of constructing custom voltage sag detection and measure-

ment subsystems. Use of equipment with advanced sag detection and measurement

capabilities would provide additional benefits by increasing the ability to respond

to sags greater than the 450ms boundary of this study, sags of varying magnitude,

and successive sags such as those seen during recloser operation. Implementation of

advanced monitoring functionality may also allow the Ridethrough PC to perform

multiple decision-making operations rather than respond to a single event at a time.

For this modification to be successfully introduced, several equipment character-

istics must be considered. Voltage sag detection and measurement functionality must

be comparable or superior to the methods shown in this study in their sensitivity,

134

accuracy, and response times. Effects of device placement on voltage measurements

must also be considered. A general preference should exist for devices in close proxim-

ity to the process under analysis to minimize compensation related to power system

configuration and connection schemes. Additionally, locally placed devices are favor-

able because they are more likely to respond to locally generated voltage sags (as with

high current motor starting) and represent fewer challenges in establishing a commu-

nication link to a software mitigation system. Network communication capabilities

are also of concern. In order for detection and measurement data to initiate response

identification and mitigation routines in a timely fashion, communication latency be-

tween the metering/protective relaying device and the mitigation system should be

minimal. Other specifications such as network protocol compatibility, network traffic

burdens, and communication methods (master/slave or peer-to-peer) should also be

assessed to determine the feasibility of applying commercially available hardware.

Microprocessor-based power quality or system protection relays offered by man-

ufacturers such as Schweitzer Engineering Laboratories, GE, ABB, or Siemens are

potentially suitable for such an application. For some equipment offerings, system

power quality information, including voltage sag data, is gathered and may be quickly

accessed from system data registers [74,75]. These devices are typically furnished

with network interfaces such as DNP, Modbus, Modbus Plus, Ethernet, or Profibus

(Siemens) that can facilitate communications with external systems. If supported by

the device, time synchronization technologies such as synchrophasor measurements

[76,77] may be employed as a means of accounting for data transfer delays. With

these features available, application of commercially available protection or metering

equipment may be realized in a software-based process voltage sag mitigation system.

Response Identification and Mitigation Routines

Both methods of identifying an expected process response are applied in an envi-

ronment that assumes either a completed process voltage sag history or knowledge of

the expected process response boundaries. Future development of response identifica-

tion should focus on determining an expected response when only a sparse data set,

135

either in the historical record or in response region boundary estimation, is available.

In this scenario, a combination of process history and boundary estimation can be

used, where a more clearly defined outcome is given a greater weight in suggesting

the expected response. Improved process response history management may be im-

plemented in an effort to fill the historical record with sags that are of the greatest

available duration, so that for equal magnitude and phasing combinations, recorded

effects under longer sags overwrite those of shorter ones. Combining these methods

may also permit automatic generation and adaptation of Boolean response identifi-

cation equations from the process history. On a system database level, traditional

SCADA process historian systems may be implemented in future studies. By ac-

cessing a global process event history, a condensed voltage sag event history may be

assembled for use in anticipated response identification. This approach can ultimately

eliminate the task of creating a separate process sag response history.

Specific mitigation algorithms may be further developed in several ways. Coast

algorithms may be improved by further optimizing algorithm response times or elim-

inating the coast state altogether by incorporating intelligent management of stored

energy injection or kinetic buffering/inertia ridethrough methods. Future signal sub-

stitution methods should account for variables that change over the course of a process

run such as package accumulation or setpoint profiling. Hybrid routines that simulta-

neously address AC drive dropout and faulted instrumentation are also possible, and

are likely to be necessary in cases where the AC drive dropout response is completely

mitigated but a faulted instrumentation response remains.

Simulation and Experimentation

Improvements in process simulation methods may involve more detailed modeling

of hardware sag responses, with the goal of entering sag characteristics or recorded

fault waveforms as inputs to create a simulated process response. By doing so, sim-

ulated results may be added to a process sag response history. If computational

expense will not affect system performance, the process dynamic model may also be

used to predict the voltage sag response to eliminate reliance on event history access.

136

Experimental verification may also be enhanced by the application of the methods

proposed in this study to a full scale industrial process. In its present or supple-

mented form, implementation and application to a variety of automation networks,

equipment, and processes can further demonstrate the advantages and limitations of

software-based voltage sag mitigation solutions.

137

APPENDICES

138

Appendix A

Textile Tension Controller Equipment List

This appendix contains tables of hardware used on the textile tension control test

stand and software used for programming and interface purposes. Ridethrough PC

equipment is not included in this Appendix.

Table A.1 Tension control system hardware information.

ITEM DESCRIPTION MANUFAC. PART OR ID NO.Load cell tension transducer Electromatic FLN-14-50-6Tension signal conditioner Electromatic TI-17A-800Pulse tach transducer Electro-Sensors 255, 906

Pulse tach signal conditioner Electro-Sensors SA420Tachometer generators Servo-Tech SB740A-7Discrete photo sensors Banner Q45BB6DProfibus I/O base Siemens IM151-1Analog input block Siemens 2AI UDiscrete input block Siemens 2DI DC24VI/O power supply Siemens PM-E DC24V

24VDC power supply Siemens PS307 5A24VDC power supply Siemens PS307 2A

S7-300 PLC/Profibus master Siemens CPU 315-2DPASI bus master Siemens CP 342-2

ASI power converter Siemens 6EP1632-1AL01ASI interface block Siemens ZU. NR. 17101

Micromaster Vector AC drive Siemens 6SE3213-6CA40Profibus drive interface Siemens CB15

1HP 3ph. duty master ind. motor Reliance S2000 - P56H0441PSerial PC/PLC interface Siemens SVPL 0303891Profibus/PC interface card Applicom PCI 1500S7

Table A.2 Programming and monitoring software.

SOFTWARE VERSIONMicrosoft Windows 98 Second Ed.

National Instruments LabView 5.1Siemens Step 7 NA

National Instruments Applicom drivers 1.1TMG-Itec Scope Profibus 5.1

139

Appendix B

Textile Tension Controller Mechanical Analysis

The mechanical portion of the tension control system consists of two drive reels

and three idle reels (Figure B.1). For analysis, each reel is treated as a rolling mass

system, and each connecting segment of textile is treated as a spring-damper system

that exerts forces on the reels it connects to. Of critical importance in this analysis is

the assumption that no idler reel slippage or web slack conditions occur in the system.

If slack or slippage were to occur, the spring-damper and rolling mass systems would

decouple, causing the system equations to become nonlinear. The constants and

variables used in this analysis are graphically shown in Figure B.2, and listed in

Table B.1. The analysis will determine the matrices for the state space relationship

x = Ax+Bu (B.1)

and

y = Cx+Du, (B.2)

where

u =τapp1τapp5

(B.3)

and

y =

⎡⎢⎢⎢⎣vtvpfpfs

⎤⎥⎥⎥⎦ . (B.4)

140

PAYOFFREEL

TAKEUPREEL

LOADCELL

IDLER

MATERIAL TRAVEL

LOADCELL

REEL 1

REEL 2

REEL 3

REEL 4

REEL 5

Figure B.1 Mechanical configuration of web tension control system.

MATERIAL TRAVEL

REEL 1B1, J1

REEL 2B2, J2

REEL 3B3, J3

REEL 4B4, J4

REEL 5B5, J5

C12

K12

C45

K45

C23

K23

C34

K34

r1

r2

r5

r3 r4

+θ1,τapp1 +θ5,τapp5

+θ2 +θ3 +θ4

Figure B.2 Mechanical constants for web tension control system.

Table B.1 Tension control system mechanical constants.

CONSTANTS DESCRIPTION UNITSθ1, θ2, θ3, θ4, θ5 reel angles unitless

θ1, θ2, θ3, θ4, θ5 reel angular velocities 1/ sec

θ1, θ2, θ3, θ4, θ5 reel angular accelerations 1/ sec2

τapp1, τapp5 drive reel applied torques N ·mB1, B2, B3, B4, B5 reel rolling frictions (kg ·m2)/ secC12, C23, C34, C45 textile damping constants (N · sec)/mfp, fs textile tensions NJ1, J2, J3, J4, J5 reel inertias kg ·m2

K12, K23, K34, K45 textile spring constants N/mr1, r2, r3, r4, r5 reel radii mv1, v5 line speeds m/ sec

141

Dynamic Analysis

We begin by evaluating the dynamic conditions for the takeup reel. Newton’s

Second Law is

Force = mass · acceleration (B.5)

and for a system in dynamic equilibrium [70],

Forces = 0. (B.6)

Since these are rolling mass systems, we sum the torques on each reel and set the

sum equal to zero. Figure B.3 graphically shows all of the torques for the takeup reel

(#1). The dynamic equation is therefore

0 = τapp1 −B1θ1 − J1θ1 − r1K12(r1θ1 − r2θ2)− r1C12(r1θ1 − r2θ2). (B.7)

REEL 1

r1

τapp1REFERENCE+θ1

B1θ1, J1θ1, r1K12(r1θ1−r2θ2), r1C12(r1θ1−r2θ2). .. . .

Figure B.3 Free body diagram of forces on reel #1 (takeup).

The idler reels are similar in their form. Figure B.4 graphically shows the torques

that contribute to the dynamic behavior for reel #2. Again we sum the torques to

obtain

0 = r2K12(r1θ1 − r2θ2) + r2C12(r1θ1 − r2θ2)−B2θ2 − J2θ2−r2K23(r2θ2 − r3θ3)− r2C23(r2θ2 − r3θ3). (B.8)

142

The equations for rollers #3 and #4 are the same as Equation B.8, with the exception

of subscript differences, which gives

0 = r3K23(r2θ2 − r3θ3) + r3C23(r2θ2 − r3θ3)−B3θ3 − J3θ3−r3K34(r3θ3 − r4θ4)− r3C34(r3θ3 − r4θ4) (B.9)

and

0 = r4K34(r3θ3 − r4θ4) + r4C34(r3θ3 − r4θ4)−B4θ4 − J4θ4−r4K45(r4θ4 − r5θ5)− r4C45(r4θ4 − r5θ5). (B.10)

The payoff reel (#5) free body diagram is shown in Figure B.5. Summing these

torques yields the equation

0 = r5K45(r4θ4 − r5θ5) + r5C45(r4θ4 − r5θ5)−B5θ5 − J5θ5 + τapp5. (B.11)

REEL 2

r2

REFERENCE+θ2

B2θ2, J2θ2, r2K23(r2θ2−r3θ3), r2C23(r2θ2−r3θ3)

r2K12(r1θ1−r2θ2),

r2C12(r1θ1−r2θ2)

.. ...

. .

Figure B.4 Free body diagram of forces on reel #2 (idler).

143

REEL 5

r5

REFERENCE+θ5

B5θ5, J5θ5

τapp5 ,

r5K45(r4θ4−r5θ5),

r5C45(r4θ4−r5θ5)

...

. .

Figure B.5 Free body diagram of forces on reel #5 (payoff).

Forming the State Equations

The state equations are formed by rearranging Equations B.7 through B.11 for

the highest order derivative, which yields

θ1 =−r21K12

J1θ1 +

−r21C12 −B1J1

θ1

+r1r2K12

J1θ2 +

r1r2C12J1

θ2 +1

J1τapp1, (B.12)

θ2 =r1r2K12

J2θ1 +

r1r2C12J2

θ1 +−r22K12 − r22K23

J2θ2

+−r22C12 − r22C23 −B2

J2θ2 +

r2r3K23

J2θ3 +

r2r3C23J2

θ3, (B.13)

θ3 =r2r3K23

J3θ2 +

r2r3C23J3

θ2 +−r23K23 − r23K34

J3θ3

+−r23C23 − r23C34 −B3

J3θ3 +

r3r4K34

J3θ4 +

r3r4C34J3

θ4, (B.14)

θ4 =r3r4K34

J4θ3 +

r3r4C34J4

θ3 +−r24K34 − r24K45

J4θ4

+−r24C34 − r24C45 −B4

J4θ4 +

r4r5K45

J4θ5 +

r4r5C45J4

θ5, (B.15)

and

θ5 =r4r5K12

J5θ4 +

r4r5C45J5

θ4

+−r25K45

J5θ5 +

−r25C45 −B5J5

θ5 +1

J5τapp5. (B.16)

144

These equations may be represented in matrix form, where the state vector is

x = θ1 θ1 θ2 θ2 θ3 θ3 θ4 θ4 θ5 θ5T, (B.17)

the state derivative vector is

x = θ1 θ1 θ2 θ2 θ3 θ3 θ4 θ4 θ5 θ5T, (B.18)

and the input vector is

u =τapp1τapp5

. (B.19)

The state matrix is formed from the state variable coefficients of equations B.12

through B.16, and by recognizing the derivative equalities amongst the elements of

Equations B.17 and B.18. Assembling this matrix gives us

A =

⎡⎢⎢⎢⎢⎢⎢⎢⎢⎢⎢⎢⎢⎢⎢⎢⎢⎢⎢⎢⎢⎢⎣

0 1 0 0 0−r21K12

J1

−r21C12−B1J1

r1r2K12

J1r1r2C12J1

0

0 0 0 1 0r1r2K12

J2r1r2C12J2

−r22K12−r22K23

J2

−r22C12−r22C23−B2J2

r2r3K23

J2

0 0 0 0 0

0 0 r2r3K23

J3r2r3C23J3

−r23K23−r23K34

J3

0 0 0 0 0

0 0 0 0 r3r4K34

J4

0 0 0 0 00 0 0 0 0

...

...

0 0 0 0 00 0 0 0 00 0 0 0 0

r2r3C23J2

0 0 0 0

1 0 0 0 0−r23C23−r23C34−B3

J3r3r4K34

J3r3r4C34J3

0 0

0 0 1 0 0r3r4C34J4

−r24K34−r24K45

J4

−r24C34−r24C45−B4J4

r4r5K45

J4r4r5C45J4

0 0 0 0 1

0 r4r5K45

J5r4r5C45J5

−r25K45

J5

−r25C45−B5J5

⎤⎥⎥⎥⎥⎥⎥⎥⎥⎥⎥⎥⎥⎥⎥⎥⎥⎥⎥⎥⎥⎥⎦

.

(B.20)

145

The input matrix is formed from the applied torque coefficients of Equations A.12

and A.16. All other values are zero, which when assembled is

B =

⎡⎣ 0 1J1

0 0 0 0 0 0 0 0

0 0 0 0 0 0 0 0 0 1J5

⎤⎦T . (B.21)

Next we form an expression to describe the process outputs of line speed and

tension in state space terms. Each process output will be computed by two avenues

to accurately reflect the redundant instrumentation found on the tension controller.

The output matrix contains line speed measured from both payoff and takeup reels,

and web tension measured by idler reels #3 and #4. Again, this matrix is

y =

⎡⎢⎢⎢⎣vtvpfpfs

⎤⎥⎥⎥⎦ . (B.22)

Line speed may be expressed as the product of a reel’s angular velocity times the

radius. For the takeup and payoff reels, this is

vt = r1θ1 (B.23)

and

vp = r5θ5. (B.24)

Each tension cell measures the average of the tension in the approaching and retreat-

ing web segments. In terms of the state variables, this is

fp =[K23(r2θ2 − r3θ3) + C23(r2θ2 − r3θ3)] + [K34(r3θ3 − r4θ4) + C34(r3θ3 − r4θ4)]

2(B.25)

and

fs =[K34(r3θ3 − r4θ4) + C34(r3θ3 − r4θ4)] + [K45(r4θ4 − r5θ5) + C45(r4θ4 − r5θ5)]

2(B.26)

Again we combine state variables, giving us

fp =r2K23

2θ2 +

r2C232

θ2 +−r3K23 + r3K34

2θ3

+−r3C23 + r3C34

2θ3 +

−r4K34

2θ4 +

−r4C342

θ4 (B.27)

146

and

fs =r3K34

2θ3 +

r3C342

θ3 +−r4K34 + r4K45

2θ4

+−r4C34 + r4C45

2θ4 +

−r5K45

2θ5 +

−r5C452

θ5. (B.28)

Finally, Equations B.23, B.24, B.27, and B.28 are combined into the matrix form

y = Cx+Du (B.29)

where

C =

⎡⎢⎢⎢⎢⎣0 r1 0 0 0 00 0 0 0 0 0

0 0 r2K23

2r2C232

−r3K23+r3K34

2−r3C23+r3C34

2

0 0 0 0 r3K34

2r3C342

...

...

0 0 0 00 0 0 r5

−r4K34

2−r4C34

20 0

−r4K34+r4K45

2−r4C34+r4C45

2−r5K45

2−r5C452

⎤⎥⎥⎥⎥⎦ , (B.30)

D =

⎡⎢⎢⎢⎣0 00 00 00 0

⎤⎥⎥⎥⎦ , (B.31)

and the matrices y, x, and u are defined in Equations B.22, B.17, and B.19, respec-

tively.

The state matrices that include values of r1 and r5 are time variant, since r1 and r5

vary throughout the unwinding and winding process. For the purpose of our analysis,

however, their values are considered to change negligibly during the timespan of sag

simulation. Therefore they are held at fixed values during simulation, causing the

state space model to remain linear and time invariant. Experimentation with the

tension control stand accounts for this as well. Voltage sags are applied during the

same point in the wind/unwind process, keeping the payoff and takeup radii consistent

for all tests.

147

Appendix C

Determination of Physical Constants

This appendix details the procedures used to determine the physical constants in

the textile control stand.

Reel Radii

Simple calipers were used to measure the reel diameters. Table C.1 shows the

resulting radii measurements.

Table C.1 Measured roller radii.

CONSTANT DESCRIPTION VALUEr1 takeup reel radius with accumulated web 0.072 mr1base takeup reel core radius 0.038 mr2 idler reel radius 0.038 mr3 primary tension cell radius 0.044 mr4 secondary tension cell radius 0.044 mr5 payoff reel radius with accumulated web 0.072 mr5base payoff reel core radius 0.038 m

Reel Inertia

To measure reel inertia, we first refer to the standard motor-load rotating system

equation

τapplied = Jdω

dt+Bω + τload. (C.1)

We may experimentally isolate the inertia component from the system by removing

the load torque, τload, and by experimenting at low rotational speeds to keep the

rolling friction component, Bω, negligible. This gives us

τapplied = Jdω

dt. (C.2)

By using calibrated weights to apply a constant torque of known value, we may

measure the rate of change of the system’s rotational speed. With these values known,

the value of J is easily solved for.

148

The final value of J is determined by assembling the results of n test runs into a

set of matrices in the form

y = Ax (C.3)

where

y =

⎡⎢⎢⎢⎢⎢⎢⎢⎣

τapplied,1τapplied,2τapplied,3...

τapplied,n

⎤⎥⎥⎥⎥⎥⎥⎥⎦ , (C.4)

A =

⎡⎢⎢⎢⎢⎢⎢⎢⎢⎣

dωdt ,1dωdt ,2dωdt ,3...

dωdt ,n

⎤⎥⎥⎥⎥⎥⎥⎥⎥⎦, (C.5)

and

x = [J ] . (C.6)

The value of x = [J ] is found using an overdetermined case solution, as outlined in

[71]. This solution is found with the equation

x = (ATA)−1ATy. (C.7)

The idler and tension reels do not have speed measurement instruments directly

attached to them, and may be referred to as being ‘unmeasurable’. Therefore, to

determine the inertia of an unmeasurable reel, a belt was attached between the un-

measurable reel and the nearest measurable (takeup or payoff) reel. Constant torques

were again applied to the measurable reel, and the two-roller system inertia was cal-

culated using the overdetermined case solution. Since the measurable reel inertia is

known from previous experiments, the unmeasurable reel inertia is the only unknown

quantity in the expression

J2rollersystem = Jmeasurable + Junmeasurablermeasurablerunmeasurable

2

, (C.8)

149

which is rearranged to give

Junmeasurable =J2rollersystem − Jmeasurable

rmeasurablerunmeasurable

2 . (C.9)

Table C.2 summarizes the results obtained from these procedures.

Table C.2 Bare roller inertias.

CONSTANT DESCRIPTION VALUEJ1base takeup reel inertia without accumulated web 0.0140 kg ·m2

J2 idler reel inertia 0.0040 kg ·m2

J3 primary tension cell inertia 0.0082 kg ·m2

J4 secondary tension cell inertia 0.0089 kg ·m2

J5base payoff reel inertia without accumulated web 0.0126 kg ·m2

Additive Inertia of Coiled Web

Added to the base inertia of the payoff and takeup reels is the inertia of the

coiled web. This may be expressed as a function of the reel radius, including the

accumulated web. The inertia of the accumulated web is the difference between the

inertia of a coiled web of radius rweb and the inertia of the theoretical center of the

coiled web whose space is take up by the reel core. This is expressed by

Jaccumulatedweb,r=rweb = Jcoiledweb,r=rweb − Jcoiledweb,r=rbase. (C.10)

To determine an expression for Equation C.10 in terms of radius, we begin with the

mass, m of a tightly coiled web, which is

m = ρπr2webw, (C.11)

where the density ρ, and width w are

ρ = 216.6kg

m3(C.12)

and

w = 0.185 m (C.13)

150

respectively. Using these values, we calculate the inertia of the theoretical center to

be

Jcoiledweb,r=rbase =1

2mr2 =

1

2ρπr2basewr

2base = 0.000262 kg ·m2 (C.14)

Similarly, the mass of the solid web cylinder of radius rweb is

Jcoiledweb,r=rweb =1

2mr2 =

1

2ρπr2webwr

2web = 62.943r

4web kg ·m2 (C.15)

Substituting values from C.14 and C.15 into C.10 yields

Jaccumulatedweb,r=rweb = (62.943r4web − 0.000262) kg ·m2 (C.16)

Equation C.16 is evaluated for r1, then added to J1base to obtain the total takeup reel

inertia, J1. The same operation is performed for the payoff reel to determine J5.

Rolling Friction

Having calculated the inertia J for each reel, we again refer to Equation C.1 for the

rolling mass system. For higher speeds the component Bω cannot be neglected. To

determine B, we observe the deceleration characteristics from an appreciable speed

under no load and no applied torque conditions. The time response takes on the form

ω(t) = ω0e−BJt (C.17)

where the value B/J is the inverse of the decay time constant, and ω0 is the initial

speed. The time constant τ was measured by curve fitting the deceleration response,

and B was calculated using

B =J

τ. (C.18)

For the idler and tension reels, a belt was again attached between an ‘unmeasur-

able’ roller and a roller with measurable speed. The two reel system’s rolling friction

constant was determined, and the unknown rolling friction constant was calculated

with

Bunmeasurable =B2rollersystem −Bmeasurable

rmeasurablerunmeasurable

2 (C.19)

which is similar in form to Equation C.9. Table C.3 shows the measured rolling

friction constants.

151

Table C.3 Rolling friction constants.

CONSTANT DESCRIPTION VALUEB1 takeup reel rolling friction 0.0018 (kg ·m2)/ secB2 idler reel rolling friction 0.0008 (kg ·m2)/ secB3 primary tension cell rolling friction 0.0022 (kg ·m2)/ secB4 secondary tension cell rolling friction 0.0023 (kg ·m2)/ secB5 payoff reel rolling friction 0.0015 (kg ·m2)/ sec

Textile Constants

The textile spring and damper constants were measured using an Instron tensile

test machine in accordance with ASTM 5035 [72]. The tensile test machine grips

onto a segment of textile and slowly pulls it apart at a constant crosshead speed.

During this operation, the force exerted by the textile on the grips is measured and

recorded as a force vs. elongation curve. From this curve, the spring constant K can

be determined as the rate of change of force per unit elongation. The value of the

spring constant was determined for several test runs at varying crosshead speeds, and

the results averaged.

The damper constant, however, was not clearly evident. If an appreciable damper

constant were present, then an appreciable initial force would be exerted at the time of

initial extension. This was not the case. Since the damper constant offers a negligible

contribution to the force exerted by the textile, it would seem reasonable to neglect

it in modeling and simulation. If it is neglected, then a high frequency component

exists in tension waveforms during simulation. Therefore, the damper constant C

was assigned a value of 1% of the spring constant, which is negligible when compared

to the spring constant, but nonzero to damp high frequency tension oscillations in

simulation. Table C.4 lists the textile spring and damper constants.

Table C.4 Textile constants.

CONSTANT DESCRIPTION VALUEK12 = K23 = K34 = K45 textile spring constants 79010 N/mC12 = C23 = C34 = C45 textile damper constants 790.10 N/(m · sec)

152

First Order Device Responses

The first order input-output relationships were determined for the analog instru-

mentation, A/D converter, and drive frequency setpoint. These relationships are

described in Equations 2.1 through 2.3. For each device, a series of steady state

outputs were determined for corresponding input, and formed into the matrices

y = Ax (C.20)

where

y =

⎡⎢⎢⎢⎢⎢⎢⎢⎣

output1output2output3...

outputn

⎤⎥⎥⎥⎥⎥⎥⎥⎦ , (C.21)

A =

⎡⎢⎢⎢⎢⎢⎢⎢⎣

input1 1input2 1input3 1...

...inputn 1

⎤⎥⎥⎥⎥⎥⎥⎥⎦ , (C.22)

and

x =Kdevice

Bdevice. (C.23)

The overdetermined case solution in Equation C.7 was then calculated for each device

to yield the values for the device gain, K and offset, B. Device time constants were

determined from manufacturer step response or construction specifications. Table

C.5 lists the first order gain, time constant, and offset results.

153

Table C.5 First order device constants.

CONSTANT DESCRIPTION VALUEKtach1 tachometer 1 gain 0.1966 V/(rad/ sec)τtach1 tachometer 1 time constant 0.0266 secBtach1 tachometer 1 offset -0.0804 VKtach5 tachometer 5 gain 0.1918 V/(rad/ sec)τtach5 tachometer 5 time constant 0.0266 secBtach5 tachometer 5 offset 0.0216 VKbtach1 backup tachometer 1 gain 0.0250 V/(rad/ sec)τbtach1 backup tachometer 1 time constant 0.0006 secBbtach1 backup tachometer 1 offset -0.0016 VKbtach5 backup tachometer 5 gain 0.0253 V/(rad/ sec)τbtach5 backup tachometer 5 time constant 0.0006 secBbtach5 backup tachometer 5 offset -0.0220 VKpten primary tension cell gain 0.0989 V/Nτpten primary tension cell time constant 0.0028 secBpten primary tension cell offset 0.0746 VKsten secondary tension cell gain 0.1023 V/Nτsten secondary tension cell time constant 0.0028 secBsten secondary tension cell offset 0.2829 VKAD A/D converter gain 2765.1 bits/VτAD A/D converter time constant 0.120 secBAD A/D converter offset -4.4997 bitsKDR1 drive 1 setpoint gain 0.01943 (rad/ sec)/bitτDR1 drive 1 setpoint time constant 0.015 secBDR1 drive 1 setpoint offset 0 (rad/ sec)KDR5 drive 5 setpoint gain 0.01943 (rad/ sec)/bitτDR5 drive 5 setpoint time constant 0.015 secBDR5 drive 5 setpoint offset 0 (rad/ sec)

154

Induction Motor Tests

To determine their equivalent circuit parameters, the induction motors were tested

with a DC stator resistance test, no-load test, and locked-rotor test, as described in

[73]. The determined equivalent circuit parameters are listed in Table C.6.

Table C.6 Induction motor equivalent circuit parameters.

CONSTANT DESCRIPTION VALUERs1 stator resistance, motor 1 3.02 ΩRr1 rotor resistance, motor 1 1.88 ΩLls1 stator leakage inductance, motor 1 0.006477 HLlr1 rotor leakage inductance, motor 1 0.009713 HLm1 magnetizing inductance, motor 1 0.181 HLs1 = Lls1 + Lm1 total stator inductance, motor 1 0.187477 HLr1 = Llr1 + Lm1 total rotor inductance, motor 1 0.190713 HRs5 stator resistance, motor 5 2.85 ΩRr5 rotor resistance, motor 5 1.92 ΩLls5 stator leakage inductance, motor 5 0.007207 HLlr5 rotor leakage inductance, motor 5 0.010810 HLm5 magnetizing inductance, motor 5 0.181 HLs5 = Lls5 + Lm5 total stator inductance, motor 5 0.188207 HLr5 = Llr5 + Lm5 total rotor inductance, motor 5 0.191810 H

155

V/Hz Drive Behavior

The V/Hz output relationship of the motor drives may be modeled in two segments

by the functions

Vout = Vebaseemef , (f < 6 Hz) (C.24)

Vout = mlf + Voffset, (f ≥ 6 Hz) (C.25)

which accounts for the low frequency voltage boost, followed by linear V/Hz output.

The relationship of output voltage versus output frequency was measured and plotted

for each drive, and the resulting curves fitted to the model in Equation C.24 using

the values listed in Table C.7.

Table C.7 Volts/Hertz drive constants.

CONSTANT DESCRIPTION VALUEVbase1 low freq. base voltage, drive 1 26.0 VRMSme1 low freq. exponential multiplier, drive 1 0.0541 1/HzVoffset1 linear segment voltage offset, drive 1 3.4799 VRMSml1 linear segment V/Hz gain, drive 1 15.0871 V/HzVbase5 low freq. base voltage, drive 5 22.5 VRMSme5 low freq. exponential multiplier, drive 5 0.0728 1/HzVoffset5 linear segment voltage offset, drive 5 3.3971 VRMSml5 linear segment V/Hz gain, drive 5 14.4466 V/Hz

156

Appendix D

Matlab Process Response Simulation Code

This appendix lists the Matlab code used to simulate the textile tension controller’s

behavior in the presence of voltage sags.

%--------------------------------------------------------------------%%sagsim.m%%Simulates the output response of a textile tension controller when%subjected to voltage sags.%%Written by Owen Parks%Clemson University PQIA Laboratories%--------------------------------------------------------------------%

%--------------------------------------------------------------------%%initialize memoryclear all;%--------------------------------------------------------------------%

%--------------------------------------------------------------------%%load sag dataload instsagdata;%101000;

%sag recovery datatau_rec=rectime/5;tau_rec2=rectime2/5;tau_rec3=rectime3/5;%--------------------------------------------------------------------%

%--------------------------------------------------------------------%%define time parameters

%evaluation points/secptspersec=10000;

%simulation durationnumsecs=3;

%total calculation points (+1 for t=0)numpts=(ptspersec*numsecs)+1;

%time arraytime=linspace(0,numsecs,numpts);

%time differentialdt=time(3)-time(2);

%sag inception timet_sagstart=0.5;

157

%sag extinction timet_sagend=0.5+0.050;

%dt for sampled waveform used in sag simulationdt2=0.1/250;%--------------------------------------------------------------------%

%--------------------------------------------------------------------%%define measured physical constants

%radii in metersr1=0.072;r2=0.038;r3=0.044;r4=0.044;r5=0.072;r1base=0.038;r5base=0.038;

%additive inertia due to accumulated webJ1web=62.943*(r1^4)-0.000262;J5web=62.943*(r5^4)-0.000262;

%inertia in kg*m^2J1=0.0140+J1web;J2=0.0040;J3=0.0082;J4=0.0089;J5=0.0126+J5web;

%rolling friction in (kg*m^2)/secB1=0.0018;B2=0.0008;B3=0.0022;B4=0.0023;B5=0.0015;

%textile spring constant in N/mK12=79010;K23=79010;K34=79010;K45=79010;

%textile damping constant in N*sec/mC12=K12*0.01;C23=K23*0.01;C34=K34*0.01;C45=K45*0.01;%--------------------------------------------------------------------%

%--------------------------------------------------------------------%%assemble state space matrices

158

%state matrixA(1,:)=[0 1 0 0 0 0 0 0 0 0];A(2,1)=(-r1*r1*K12)/(J1);A(2,2)=(-r1*r1*C12-B1)/(J1);A(2,3)=(r1*r2*K12)/(J1);A(2,4)=(r1*r2*C12)/(J1);A(2,5)=0;A(2,6)=0;A(2,7)=0;A(2,8)=0;A(2,9)=0;A(2,10)=0;A(3,:)=[0 0 0 1 0 0 0 0 0 0];A(4,1)=(r1*r2*K12)/(J2);A(4,2)=(r1*r2*C12)/(J2);A(4,3)=(-r2*r2*K12-r2*r2*K23)/(J2);A(4,4)=(-r2*r2*C12-r2*r2*C23-B2)/(J2);A(4,5)=(r2*r3*K23)/(J2);A(4,6)=(r2*r3*C23)/(J2);A(4,7)=0;A(4,8)=0;A(4,9)=0;A(4,10)=0;A(5,:)=[0 0 0 0 0 1 0 0 0 0];A(6,1)=0;A(6,2)=0;A(6,3)=(r2*r3*K23)/(J3);A(6,4)=(r2*r3*C23)/(J3);A(6,5)=(-r3*r3*K23-r3*r3*K34)/(J3);A(6,6)=(-r3*r3*C23-r3*r3*C34-B3)/(J3);A(6,7)=(r3*r4*K34)/(J3);A(6,8)=(r3*r4*C34)/(J3);A(6,9)=0;A(6,10)=0;A(7,:)=[0 0 0 0 0 0 0 1 0 0];A(8,1)=0;A(8,2)=0;A(8,3)=0;A(8,4)=0;A(8,5)=(r3*r4*K34)/(J4);A(8,6)=(r3*r4*C34)/(J4);A(8,7)=(-r4*r4*K34-r4*r4*K45)/(J4);A(8,8)=(-r4*r4*C34-r4*r4*C45-B4)/(J4);A(8,9)=(r4*r5*K45)/(J4);A(8,10)=(r4*r5*C45)/(J4);A(9,:)=[0 0 0 0 0 0 0 0 0 1];A(10,1)=0;A(10,2)=0;A(10,3)=0;A(10,4)=0;A(10,5)=0;A(10,6)=0;A(10,7)=(r4*r5*K45)/(J5);A(10,8)=(r4*r5*C45)/(J5);

159

A(10,9)=(-r5*r5*K45)/(J5);A(10,10)=(-r5*r5*C45-B5)/(J5);

%input matrixB=[0 0; (1/J1) 0; 0 0; 0 0; 0 0; 0 0; 0 0; 0 0; 0 0; 0 (1/J5)];

%output matrixC(1,:)=[0 r1 0 0 0 0 0 0 0 0];C(2,:)=[0 0 0 0 0 0 0 0 0 r5];C(3,1)=0;C(3,2)=0;C(3,3)=(r2*K23)/(2);C(3,4)=(r2*C23)/(2);C(3,5)=(-r3*K23+r3*K34)/(2);C(3,6)=(-r3*C23+r3*C34)/(2);C(3,7)=(-r4*K34)/(2);C(3,8)=(-r4*C34)/(2);C(3,9)=0;C(3,10)=0;C(4,1)=0;C(4,2)=0;C(4,3)=0;C(4,4)=0;C(4,5)=(r3*K34)/(2);C(4,6)=(r3*C34)/(2);C(4,7)=(-r4*K34+r4*K45)/(2);C(4,8)=(-r4*C34+r4*C45)/(2);C(4,9)=(-r5*K45)/(2);C(4,10)=(-r5*C45)/(2);

%feedthrough matrixD=zeros(4,2);%--------------------------------------------------------------------%

%--------------------------------------------------------------------%%define control constants

%proportional, integral, and state gain for linespeedP_linespeed=0.3025;I_linespeed=1/0.566;K_state_linespeed=0.005;

%proportional, integral, and state gain for tensionP_tension=0.3;I_tension=1/0.090;K_state_tension=0.02;%--------------------------------------------------------------------%

%--------------------------------------------------------------------%%setpoints for speed and tension - held constant

%linespeed set in m/slinespeed_set=1.0;

160

%tension set in Ntension_set=30;%--------------------------------------------------------------------%

%--------------------------------------------------------------------%%define gain and time constant for first order hardware

%takeup speed sensorK_tach1=0.1966;tau_tach1=0.0266;offset_tach1=-0.0804;

%payoff speed sensorK_tach5=0.1918;tau_tach5=0.0266;offset_tach5=0.0216;

%backup takeup speed sensorK_btach1=0.0250;tau_btach1=0.0006;offset_btach1=-0.0016;

%backup payoff speed sensorK_btach5=0.0253;tau_btach5=0.0006;offset_btach5=0.0220;

%primary tension sensorK_pten=0.0989;tau_pten=0.0028;offset_pten=0.0746;

%secondary tension sensorK_sten=0.1023;tau_sten=0.0028;offset_sten=0.2829;

%input A/D blockK_ad=2765.1;tau_ad=0.120;offset_ad=-4.4997;

%drive 1 output frequency - convert to rad/sec, not HzK_DR1=0.01943;tau_DR1=0.015;%no drive 1 output offset

%drive 5 output frequency - convert to rad/sec, not HzK_DR5=0.01943;tau_DR5=0.015;%no drive 5 output offset%--------------------------------------------------------------------%

%--------------------------------------------------------------------%

161

%define AC induction motor constants

%motor 1Rs1=2.77;Rr1=1.999;Lls1=0.00617;Llr1=0.00925;Lm1=0.1731;Ls1=Lls1+Lm1;Lr1=Llr1+Lm1;Lmat1=[(Lls1+Lm1) 0 Lm1 0; 0 (Lls1+Lm1) 0 Lm1;...

Lm1 0 (Llr1+Lm1) 0; 0 Lm1 0 (Llr1+Lm1)];

%motor 5Rs5=2.756;Rr5=1.831;Lls5=0.006903;Llr5=0.010354;Lm5=0.17070;Ls5=Lls5+Lm5;Lr5=Llr5+Lm5;Lmat5=[(Lls5+Lm5) 0 Lm5 0; 0 (Lls5+Lm5) 0 Lm5;...

Lm5 0 (Llr5+Lm5) 0; 0 Lm5 0 (Llr5+Lm5)];%--------------------------------------------------------------------%

%--------------------------------------------------------------------%%define constants for motor drives

%volts per hertz constants >= 6HzVpHzgain1=3.4799;VpHzgain5=3.3971;VpHzoffset1=15.0871;VpHzoffset5=14.4466;

%volts per hertz constants < 6HzVpHzegain1=26.0;VpHzegain5=22.5;VpHzexp1=log((VpHzgain1*6+VpHzoffset1)/VpHzegain1)/6;VpHzexp5=log((VpHzgain5*6+VpHzoffset5)/VpHzegain5)/6;

%rotor voltages - shorted for squirrel cage induction motorvqr1=0;vdr1=0;vqr5=0;vdr5=0;%--------------------------------------------------------------------%

%--------------------------------------------------------------------%%initialize integration sums and output signals at t=0

%initialize linespeed and tension outputsinitialspeed=linespeed_set;initialtension=tension_set;

162

%assign array inital valueslinespeed(1)=initialspeed;pten_actual(1)=initialtension;

%calculate initial matrix states - x and uxdot=[linespeed(1)/r1; 0; linespeed(1)/r2; 0; linespeed(1)/r3; 0;...

linespeed(1)/r4; 0; linespeed(1)/r5; 0];x=[0; linespeed(1)/r1; 0; linespeed(1)/r2; 0; linespeed(1)/r3;...

0; linespeed(1)/r4; 0; linespeed(1)/r5];

%calculate angles theta1-theta5findthetamatrix=zeros(5,5);findthetamatrix(1,2)=r2*K23;findthetamatrix(1,3)=-r3*K23+r3*K34;findthetamatrix(1,4)=-r4*K34;findthetamatrix(2,1)=r1*r2*K12;findthetamatrix(2,2)=-r2*r2*K12-r2*r2*K23;findthetamatrix(2,3)=r2*r3*K23;findthetamatrix(3,2)=r2*r3*K23;findthetamatrix(3,3)=-r3*r3*K23-r3*r3*K34;findthetamatrix(3,4)=r3*r4*K34;findthetamatrix(4,3)=r3*r4*K34;findthetamatrix(4,4)=-r4*r4*K34-r4*r4*K45;findthetamatrix(4,5)=r4*r5*K45;findthetamatrix(5,3)=1;findthetavalues=[2*initialtension; B2*x(4); B3*x(6); B4*x(8); 10*pi];findthetas=inv(findthetamatrix)*findthetavalues;

x(1)=findthetas(1);x(3)=findthetas(2);x(5)=findthetas(3);x(7)=findthetas(4);x(9)=findthetas(5);

x_intsum=x;

initialtorques=(xdot-(A*x));torque_1(1)=initialtorques(2)*J1;torque_5(1)=initialtorques(10)*J5;

yinit=C*x;

%instrument output signals at t=0; w=1thetadot_1(1)=linespeed_set/r1;out_tach1(1)=thetadot_1(1)*K_tach1+offset_tach1;out_tach1B(1)=thetadot_1(1)*K_tach1+offset_tach1;thetadot_5(1)=linespeed_set/r5;out_tach5(1)=thetadot_5(1)*K_tach5+offset_tach5;out_tach5B(1)=thetadot_5(1)*K_tach5+offset_tach5;out_btach1(1)=thetadot_1(1)*K_btach1+offset_btach1;out_btach5(1)=thetadot_5(1)*K_btach5+offset_btach5;pten_actual(1)=yinit(3);out_pten(1)=pten_actual(1)*K_pten+offset_pten;out_pten2(1)=out_pten(1);

163

sten_actual(1)=yinit(4);out_sten(1)=sten_actual(1)*K_sten+offset_sten;out_ad1(1)=out_tach1(1)*K_ad+offset_ad;out_ad2(1)=out_tach5(1)*K_ad+offset_ad;out_ad3(1)=out_pten(1)*K_ad+offset_ad;out_ad1B(1)=out_ad1(1);out_ad2B(1)=out_ad2(1);out_ad3B(1)=out_ad3(1);out_ad1A(1)=out_ad1(1);out_ad2A(1)=out_ad2(1);out_ad3A(1)=out_ad3(1);

%find linespeed integral control initial value%iterative application of dq theory required

%search above thetadot for motoringsweeppts=100000;omegaesweep=linspace(2*thetadot_1(1),2*thetadot_1(1)+20,sweeppts);q=1;

for z=1:sweepptsif abs(omegaesweep(z))<(6*2*pi)

llvolts=VpHzegain1*exp(VpHzexp1*(omegaesweep(z)/(2*pi)));voltmag=llvolts/sqrt(3);

elseif abs(omegaesweep(z))>=(6*2*pi)llvolts=VpHzgain1*(omegaesweep(z)/(2*pi))+VpHzoffset1;voltmag=llvolts/sqrt(3);

end

vas=voltmag*sqrt(2)*cos(0);vbs=voltmag*sqrt(2)*cos(0-2*pi/3);vcs=voltmag*sqrt(2)*cos(0+2*pi/3);

vqss=(2/3)*vas-(1/3)*vbs-(1/3)*vcs;vdss=(-1/sqrt(3))*vbs+(1/sqrt(3))*vcs;

vqs=vqss*cos(0)-vdss*sin(0);vds=vqss*sin(0)+vdss*cos(0);

vmat=[vqs; vds; 0; 0];initmat=[Rs1 omegaesweep(z)*Ls1 0 omegaesweep(z)*Lm1;...

-1*omegaesweep(z)*Ls1 Rs1 -1*omegaesweep(z)*Lm1 0;...0 (omegaesweep(z)-2*thetadot_1(1))*Lm1 ...Rr1 (omegaesweep(z)-2*thetadot_1(1))*Lr1;...-1*(omegaesweep(z)-2*thetadot_1(1))*Lm1 0 ...-1*(omegaesweep(z)-2*thetadot_1(1))*Lr1 Rr1];

i_init=inv(initmat)*vmat;iqs=i_init(1);ids=i_init(2);iqr=i_init(3);idr=i_init(4);testtorque=3*Lm1*(iqs*idr-ids*iqr);

if abs(testtorque-torque_1(1))<0.001

164

initialomega1(q)=omegaesweep(z);q=q+1;

end

end

%electrical frequency that yields the desired speed-torque pointomegaefound=mean(initialomega1);

%back calculate to integral outputintsum_Ilinespeed=(((omegaefound/(K_DR1*276.4))...

-K_state_linespeed*r5*yinit(3))/P_linespeed)/I_linespeed;

%find tension integral control initial value%iterative application of dq theory required

%search below thetadot for brakingsweeppts=100000;omegaesweep=linspace(2*thetadot_5(1),2*thetadot_5(1)-20,sweeppts);q=1;

for z=1:sweepptsif abs(omegaesweep(z))<(6*2*pi)

llvolts=VpHzegain5*exp(VpHzexp5*(omegaesweep(z)/(2*pi)));voltmag=llvolts/sqrt(3);

elseif abs(omegaesweep(z))>=(6*2*pi)llvolts=VpHzgain5*(omegaesweep(z)/(2*pi))+VpHzoffset5;voltmag=llvolts/sqrt(3);

end

vas=voltmag*sqrt(2)*cos(0);vbs=voltmag*sqrt(2)*cos(0-2*pi/3);vcs=voltmag*sqrt(2)*cos(0+2*pi/3);

vqss=(2/3)*vas-(1/3)*vbs-(1/3)*vcs;vdss=(-1/sqrt(3))*vbs+(1/sqrt(3))*vcs;

vqs=vqss*cos(0)-vdss*sin(0);vds=vqss*sin(0)+vdss*cos(0);

vmat=[vqs; vds; 0; 0];initmat=[Rs5 omegaesweep(z)*Ls5 0 omegaesweep(z)*Lm5;...

-1*omegaesweep(z)*Ls5 Rs5 -1*omegaesweep(z)*Lm5 0;...0 (omegaesweep(z)-2*thetadot_5(1))*Lm5 ...Rr5 (omegaesweep(z)-2*thetadot_5(1))*Lr5;...-1*(omegaesweep(z)-2*thetadot_5(1))*Lm5 0 ...-1*(omegaesweep(z)-2*thetadot_5(1))*Lr5 Rr5];

i_init=inv(initmat)*vmat;iqs=i_init(1);ids=i_init(2);iqr=i_init(3);idr=i_init(4);testtorque=3*Lm5*(iqs*idr-ids*iqr);

165

if abs(testtorque-torque_5(1))<0.001initialomega5(q)=omegaesweep(z);q=q+1;

end

end

%electrical frequency that yields the desired speed-torque pointomegaefound=mean(initialomega5);

%back calculate to integral outputintsum_Itension=(((omegaefound/(K_DR5*276.4))...

-K_state_tension*x(2))/P_tension)/I_tension;

%control signals at t=0; w=1linespeed_err(1)=0;tension_err(1)=0;linespeed_ctrl_out(1)=(P_linespeed*I_linespeed*intsum_Ilinespeed...

+K_state_linespeed*r5*yinit(3))*276.4;tension_ctrl_out(1)=(P_tension*I_tension*intsum_Itension...

+K_state_tension*x(2))*276.4;

%control integrator initializationw_integral=1;linespeed_I_cont=I_linespeed*intsum_Ilinespeed;tension_I_cont=I_tension*intsum_Itension;

%motor 1 and drive 1 values at t=0; w=1omegae1(1)=linespeed_ctrl_out(1)*K_DR1;drive1_thetas(1)=0;

if abs(omegae1(1))<(6*2*pi)llvolts=VpHzegain1*exp(VpHzexp1*(omegae1(1)/(2*pi)));voltmag_1=llvolts/sqrt(3);

elseif abs(omegae1(1))>=(6*2*pi)llvolts=VpHzgain1*(omegae1(1)/(2*pi))+VpHzoffset1;voltmag_1=llvolts/sqrt(3);

end

vas1=voltmag_1*sqrt(2)*cos(0);vbs1=voltmag_1*sqrt(2)*cos(0-2*pi/3);vcs1=voltmag_1*sqrt(2)*cos(0+2*pi/3);

vqss1=(2/3)*vas1-(1/3)*vbs1-(1/3)*vcs1;vdss1=(-1/sqrt(3))*vbs1+(1/sqrt(3))*vcs1;

vqs1(1)=vqss1*cos(0)-vdss1*sin(0);vds1(1)=vqss1*sin(0)+vdss1*cos(0);

vmat1=[vqs1(1); vds1(1); vqr1; vdr1];initmat_1=[Rs1 omegae1(1)*Ls1 0 omegae1(1)*Lm1;...

-1*omegae1(1)*Ls1 Rs1 -1*omegae1(1)*Lm1 0;...0 (omegae1(1)-2*thetadot_1(1))*Lm1 ...Rr1 (omegae1(1)-2*thetadot_1(1))*Lr1;...

166

-1*(omegae1(1)-2*thetadot_1(1))*Lm1 0 ...-1*(omegae1(1)-2*thetadot_1(1))*Lr1 Rr1];

i_init1=inv(initmat_1)*vmat1;iqs1(1)=i_init1(1);ids1(1)=i_init1(2);iqr1(1)=i_init1(3);idr1(1)=i_init1(4);

psi_init1=Lmat1*i_init1;psi_qs1(1)=psi_init1(1);psi_ds1(1)=psi_init1(2);psi_qr1(1)=psi_init1(3);psi_dr1(1)=psi_init1(4);

%motor 5 and drive 5 values at t=0; w=1omegae5(1)=tension_ctrl_out(1)*K_DR5;drive5_thetas(1)=0;

if abs(omegae5(1))<(6*2*pi)llvolts=VpHzegain5*exp(VpHzexp5*(omegae5(1)/(2*pi)));voltmag_5=llvolts/sqrt(3);

elseif abs(omegae5(1))>=(6*2*pi)llvolts=VpHzgain5*(omegae5(1)/(2*pi))+VpHzoffset5;voltmag_5=llvolts/sqrt(3);

end

vas5=voltmag_5*sqrt(2)*cos(0);vbs5=voltmag_5*sqrt(2)*cos(0-2*pi/3);vcs5=voltmag_5*sqrt(2)*cos(0+2*pi/3);


vqs5(1)=vqss5*cos(0)-vdss5*sin(0);vds5(1)=vqss5*sin(0)+vdss5*cos(0);

vmat5=[vqs5(1); vds5(1); vqr5; vdr5];initmat_5=[Rs5 omegae5(1)*Ls5 0 omegae5(1)*Lm5;...

-1*omegae5(1)*Ls5 Rs5 -1*omegae5(1)*Lm5 0;...0 (omegae5(1)-2*thetadot_5(1))*Lm5 ...Rr5 (omegae5(1)-2*thetadot_5(1))*Lr5;...-1*(omegae5(1)-2*thetadot_5(1))*Lm5 0 ...-1*(omegae5(1)-2*thetadot_5(1))*Lr5 Rr5];

i_init5=inv(initmat_5)*vmat5;iqs5(1)=i_init5(1);ids5(1)=i_init5(2);iqr5(1)=i_init5(3);idr5(1)=i_init5(4);

psi_init5=Lmat5*i_init5;psi_qs5(1)=psi_init5(1);psi_ds5(1)=psi_init5(2);

167

psi_qr5(1)=psi_init5(3);psi_dr5(1)=psi_init5(4);

%back calculate initial integral values - several use w=1 valuesintsum_tach1=out_tach1(1);intsum_tach5=out_tach5(1);intsum_btach1=out_btach1(1);intsum_btach5=out_btach5(1);intsum_pten=out_pten(1);intsum_sten=out_sten(1);intsum_ad1=out_ad1(1);intsum_ad2=out_ad2(1);intsum_ad3=out_ad3(1);

intsum_omegae1=omegae1(1);drive1_thetas_int=0;psi_qs1_int=psi_qs1(1);psi_ds1_int=psi_ds1(1);psi_qr1_int=psi_qr1(1);psi_dr1_int=psi_dr1(1);

intsum_omegae5=omegae5(1);drive5_thetas_int=0;psi_qs5_int=psi_qs5(1);psi_ds5_int=psi_ds5(1);psi_qr5_int=psi_qr5(1);psi_dr5_int=psi_dr5(1);

%create arrays for signal substitution and control substitutiontension_replace(1:numpts)=0;tach1_replace(1:numpts)=0;tach5_replace(1:numpts)=0;control_substitution(1:numpts)=0;

display(’INITIAL CONDITIONS SET’)clock%--------------------------------------------------------------------%

%--------------------------------------------------------------------%%instrument sag simulation - reset variables for recoveryintreset=1;intreset2=1;intreset3=1;%--------------------------------------------------------------------%

%--------------------------------------------------------------------%%indices for referencing sampled wavesq=0;r=1;q2=0;r2=1;q3=0;r3=1;%--------------------------------------------------------------------%

168

%--------------------------------------------------------------------%%drive variables for trip / no tripdcbus1(1:numpts)=280;dcbus5(1:numpts)=280;dcdecayA=975; %v/sec dc bus decay rate: energized motordcdecayB=46.6; %v/sec dc bus decay rate: nonenergized motordrivetrip1=199.5;drivetrip5=199.5;trip1=0;trip5=0;driveout1(1:numpts)=1;driveout5(1:numpts)=1;coastcommand1(1:numpts)=0;coastcommand5(1:numpts)=0;%--------------------------------------------------------------------%

%--------------------------------------------------------------------%%enter loop for each point in time and run calculationsfor w=2:numpts %loop for number of evaluation points

%----------------------------------------------------------------%%feedback instruments - all first order%----------------------------------------------------------------%%main tachometer - roller 1%input: thetadot_1(w-1)%output: out_tach1B(w)

%standard operation - prefault, ridethrough, and postfault%standard first order for instrumenttach1_integral=dt*((K_tach1/tau_tach1)*thetadot_1(w-1))...

-dt*((1/tau_tach1)*(out_tach1(w-1)-offset_tach1));intsum_tach1=tach1_integral+intsum_tach1;out_tach1(w)=intsum_tach1;out_tach1B(w)=out_tach1(w);

%sagged operation - deviation segmentif time(w)>t_sagstart+ridetime2 & time(w)<=t_sagend

%routine to handle dissimilar samplingtimeindex=time(w)-time(w-q2);if timeindex < (time(w)-time(w-4))

q2=q2+1;out_tach1B(w)=wave2(r2);

elseif timeindex >= (time(w)-time(w-4))q2=1;r2=r2+1;out_tach1B(w)=wave2(r2);

endend

%sagged operation - recovery segmentif time(w)>t_sagend & time(w)<t_sagend+10*tau_rec2

%initial value setif intreset2==1

intreset2=0;

169

initval2=out_tach1B(w-1);end

%calculate gradual sweep multipliersdec_multiplier2=exp(-(time(w)-t_sagend)/tau_rec2);inc_multiplier2=1-exp(-(time(w)-t_sagend)/tau_rec2);

%signal definition - equation 2.29out_tach1B(w)=initval2*dec_multiplier2+out_tach1(w)*inc_multiplier2;

end

%----------------------------------------------------------------%%main tachometer - roller 5%input: thetadot_5(w-1)%output: out_tach5B(w)

%standard operation - prefault, ridethrough, and postfault%standard first order for instrumenttach5_integral=dt*((K_tach5/tau_tach5)*thetadot_5(w-1))...

-dt*((1/tau_tach5)*(out_tach5(w-1)-offset_tach5));intsum_tach5=tach5_integral+intsum_tach5;out_tach5(w)=intsum_tach5;out_tach5B(w)=out_tach5(w);

%sagged operation - deviation segmentif time(w)>t_sagstart+ridetime3 & time(w)<=t_sagend

%routine to handle dissimilar samplingtimeindex=time(w)-time(w-q3);if timeindex < (time(w)-time(w-4))

q3=q3+1;out_tach5B(w)=wave3(r3);

elseif timeindex >= (time(w)-time(w-4))q3=1;r3=r3+1;out_tach5B(w)=wave3(r3);

endend

%sagged operation - recovery segmentif time(w)>t_sagend & time(w)<t_sagend+10*tau_rec3

%initial value setif intreset3==1

intreset3=0;initval3=out_tach5B(w-1);

end

%calculate gradual sweep multipliersdec_multiplier3=exp(-(time(w)-t_sagend)/tau_rec3);inc_multiplier3=1-exp(-(time(w)-t_sagend)/tau_rec3);

%signal definition - equation 2.29out_tach5B(w)=initval3*dec_multiplier3+out_tach5(w)*inc_multiplier3;

end

170

%----------------------------------------------------------------%%backup tachometer - roller 1%input: thetadot_1(w-1)%output: out_btach1(w)btach1_integral=dt*((K_btach1/tau_btach1)*thetadot_1(w-1))...

-dt*((1/tau_btach1)*(out_btach1(w-1)-offset_btach1));intsum_btach1=btach1_integral+intsum_btach1;out_btach1(w)=intsum_btach1;

%backup tachometer - roller 5%input: thetadot_5(w-1)%output: out_btach5(w)btach5_integral=dt*((K_btach5/tau_btach5)*thetadot_5(w-1))...

-dt*((1/tau_btach5)*(out_btach5(w-1)-offset_btach5));intsum_btach5=btach5_integral+intsum_btach5;out_btach5(w)=intsum_btach5;

%----------------------------------------------------------------%%primary tension cell%input: pten_actual(w-1)%output: out_pten2(w)

%standard operation - prefault, ridethrough, and postfault%standard first order for instrumentpten_integral=dt*((K_pten/tau_pten)*pten_actual(w-1))...-dt*((1/tau_pten)*(out_pten(w-1)-offset_pten));intsum_pten=pten_integral+intsum_pten;out_pten(w)=intsum_pten;out_pten2(w)=out_pten(w);

%sagged operation - deviation segmentif time(w)>t_sagstart+ridetime & time(w)<=t_sagend

%routine to handle dissimilar samplingtimeindex=time(w)-time(w-q);if timeindex < (time(w)-time(w-4))

q=q+1;out_pten2(w)=wave(r);

elseif timeindex >= (time(w)-time(w-4))q=1;r=r+1;out_pten2(w)=wave(r);

endend

%sagged operation - recovery segmentif time(w)>t_sagend & time(w)<t_sagend+10*tau_rec

%initial value setif intreset==1

intreset=0;initval=out_pten2(w-1);

end

%calculate gradual sweep multipliersdec_multiplier=exp(-(time(w)-t_sagend)/tau_rec);

171

inc_multiplier=1-exp(-(time(w)-t_sagend)/tau_rec);

%signal definition - equation 2.29out_pten2(w)=initval*dec_multiplier+out_pten(w)*inc_multiplier;

end

%----------------------------------------------------------------%%secondary tension cell%input: sten_actual(w-1)%output: out_sten(w)sten_integral=dt*((K_sten/tau_sten)*sten_actual(w-1))...

-dt*((1/tau_sten)*(out_sten(w-1)-offset_sten));intsum_sten=sten_integral+intsum_sten;out_sten(w)=intsum_sten;%----------------------------------------------------------------%

%----------------------------------------------------------------%%instrument input block - three instances

%zero order hold for samplingtimeindex1=0.1*(floor(time(w)/0.1));oldw1=floor((timeindex1/(1/10000))+1);

%tach1 input signal%input: out_tach1(w-1)%output: out_ad1(w)ad_integral1=dt*((K_ad/tau_ad)*out_tach1B(w-1))...

-dt*((1/tau_ad)*(out_ad1(w-1)-offset_ad));intsum_ad1=ad_integral1+intsum_ad1;out_ad1A(w)=intsum_ad1;out_ad1(w)=out_ad1A(oldw1);out_ad1B(w)=out_ad1(w);

%tach5 input signal%input: out_tach5(w-1)%output: out_ad2(w)ad_integral2=dt*((K_ad/tau_ad)*out_tach5B(w-1))...


%primary tension input signal%input: out_pten2(w-1)%output: out_ad3(w)ad_integral3=dt*((K_ad/tau_ad)*out_pten2(w-1))...


%INTERRUPT SIGNALS IF RIDETIMES EXCEEDED - SELECTIVE SUBSTITUTION

172

%3 IF STATEMENTS REASSIGN OUT_AD*B VALUES% if time(w)>=0.5+ridetime2 & time(w)<=0.5+0.450+rectime2% out_ad1B(w)=mean(out_ad1B(10:4000));% intsum_ad1=out_ad1B(w);% tach1_replace(w)=1;% end% if time(w)>=0.5+ridetime3 & time(w)<=0.5+0.450+rectime3% out_ad2B(w)=mean(out_ad2B(10:4000));% intsum_ad2=out_ad2B(w);% tach5_replace(w)=1;% end% if time(w)>=0.5+ridetime & time(w)<=0.5+0.450+rectime% out_ad3B(w)=mean(out_ad3B(10:4000));% intsum_ad3=out_ad3B(w);% tension_replace(w)=1;% end

%----------------------------------------------------------------%

%----------------------------------------------------------------%%control calculations

%scale feedback signals

%tach1 signaltach1_sig=(((out_ad1B(w)-offset_ad)/K_ad)-offset_tach1)/K_tach1;

%tach5 signaltach5_sig=(((out_ad2B(w)-offset_ad)/K_ad)-offset_tach5)/K_tach5;

%primary tension signalpten_sig=(((out_ad3B(w)-offset_ad)/K_ad)-offset_pten)/K_pten;

%calculate PVs from scaled feed back signalslinespeed_PV=tach1_sig;tension_PV=pten_sig*r5;

%calculate control SPs from scaled feedback signals and%desired values of linespeed and tensionlinespeed_SP=linespeed_set/r1;tension_SP=tension_set*r5;

%calculate control signal errorslinespeed_err(w)=linespeed_SP-linespeed_PV;tension_err(w)=tension_PV-tension_SP;

%calculate integral contribution to control%calculated every PLC cycledt_PLC=0.01; %PLC cycle interval

%IF STATEMENT MANAGES INTEGRAL HOLD, ELSE IS FOR NORMAL OPERATIONif time(w)>=t_sagstart+0.006 & time(w)<=t_sagend+1.5

intsum_Ilinespeed=intsum_Ilinespeed;intsum_Itension=intsum_Itension;

173

elseif time(w)-time(w_integral)>=dt_PLC-0.000001

w_integral=w;intsum_Ilinespeed=intsum_Ilinespeed+linespeed_err(w)*dt_PLC;linespeed_I_cont=I_linespeed*intsum_Ilinespeed;intsum_Itension=intsum_Itension+tension_err(w)*dt_PLC;tension_I_cont=I_tension*intsum_Itension;

endend

%record integral sums for referenceI_line(w)=intsum_Ilinespeed;I_ten(w)=intsum_Itension;

%calculate proportional contribution to controllinespeed_P_cont=P_linespeed*linespeed_err(w);tension_P_cont=P_tension*tension_err(w);

%sum proportional and integral termslinespeed_PIsum(w)=linespeed_P_cont+P_linespeed*linespeed_I_cont;tension_PIsum(w)=tension_P_cont+P_tension*tension_I_cont;

%calculate state feedback termslinespeed_state_cont(w)=r5*pten_sig*K_state_linespeed;tension_state_cont(w)=tach5_sig*K_state_tension;

%sum state feedback term and PI sumlinespeed_control(w)=linespeed_PIsum(w)+linespeed_state_cont(w);tension_control(w)=tension_PIsum(w)+tension_state_cont(w);

%multiply by intrinsic constant used in Siemens PLClinespeed_ctrl_out(w)=linespeed_control(w)*276.4;tension_ctrl_out(w)=tension_control(w)*276.4;%----------------------------------------------------------------%

%----------------------------------------------------------------%%drive maximum and minimum saturation at +-15Hzif linespeed_ctrl_out(w)>((15*2*pi)/K_DR1)

linespeed_ctrl_out(w)=((15*2*pi)/K_DR1);elseif linespeed_ctrl_out(w)<((-15*2*pi)/K_DR1)

linespeed_ctrl_out(w)=((-15*2*pi)/K_DR1);end

if tension_ctrl_out(w)>((15*2*pi)/K_DR5)tension_ctrl_out(w)=((15*2*pi)/K_DR5);

elseif tension_ctrl_out(w)<((-15*2*pi)/K_DR5)tension_ctrl_out(w)=((-15*2*pi)/K_DR5);

end

%CALCULATE AND OUTPUT PREFAULT AVERAGE VALUES FOR%CONTROL OUTPUTSif time(w)>=t_sagstart+0.006 & time(w)<=t_sagend+1.5

linespeed_ctrl_out(w)=mean(linespeed_ctrl_out(10:4000));tension_ctrl_out(w)=mean(tension_ctrl_out(10:4000));

174

control_substitution(w)=1;end%----------------------------------------------------------------%

%----------------------------------------------------------------%%motor drives%inputs: linespeed_ctrl_out(w),tension_ctrl_out(w)%ouputs: torque_1(w),torque_5(w)%each drive done separately

%***drive 1%determine rotating field frequencyomegae1_integral=dt*((K_DR1/tau_DR1)*linespeed_ctrl_out(w-1))...

-dt*((1/tau_DR1)*omegae1(w-1));omegae1(w)=omegae1_integral+intsum_omegae1;intsum_omegae1=omegae1(w);

%determine anglesdrive1_thetas(w)=drive1_thetas_int+dt*omegae1(w);drive1_thetas_int=drive1_thetas(w);

if abs(omegae1(w))<(6*2*pi)llvolts=VpHzegain1*exp(VpHzexp1*(omegae1(w)/(2*pi)));voltmag_1=llvolts/sqrt(3);

elseif abs(omegae1(w))>=(6*2*pi)llvolts=VpHzgain1*(omegae1(w)/(2*pi))+VpHzoffset1;voltmag_1=llvolts/sqrt(3);

end

%COAST COMMANDif time(w)>=t_sagstart+0.025 & time(w)<=t_sagend+0.002

coastcommand1(w)=1;end

%DRIVE DROPOUT SEQUENCE AND DC BUS DECAY%SECOND CONDITIONAL INCLUDED IN RECOVERYif time(w)>=t_sagstart & time(w)<=t_sagend

if dcbus1(w-1)>=drivetrip1dcbus1(w)=dcbus1(w-1)-dcdecayA*dt;

endif dcbus1(w-1)<drivetrip1 | coastcommand1(w)==1

dcbus1(w)=dcbus1(w-1)-dcdecayB*dt;driveout1(w)=0;voltmag_1=0;psi_qs1(w-1)=0;psi_qr1(w-1)=0;psi_ds1(w-1)=0;psi_dr1(w-1)=0;

endend

if trip1==0 & dcbus1(w)<drivetrip1trip1=1;

end

175

%if driveout1(w-1)==0 %maintains trip for no mitigationif trip1==1;

driveout1(w)=0;voltmag_1=0;psi_qs1(w-1)=0;psi_qr1(w-1)=0;psi_ds1(w-1)=0;psi_dr1(w-1)=0;

end

vas1=voltmag_1*sqrt(2)*cos(drive1_thetas(w));vbs1=voltmag_1*sqrt(2)*cos(drive1_thetas(w)-2*pi/3);vcs1=voltmag_1*sqrt(2)*cos(drive1_thetas(w)+2*pi/3);

vqss1=(2/3)*vas1-(1/3)*vbs1-(1/3)*vcs1;vdss1=-(1/sqrt(3))*vbs1+(1/sqrt(3))*vcs1;

vqs1(w)=vqss1*cos(drive1_thetas(w))-vdss1*sin(drive1_thetas(w));vds1(w)=vqss1*sin(drive1_thetas(w))+vdss1*cos(drive1_thetas(w));

%calculate dq fluxespsi_qs1(w)=psi_qs1_int+dt*(vqs1(w-1)-Rs1*iqs1(w-1)...

-omegae1(w-1)*psi_ds1(w-1));psi_qs1_int=psi_qs1(w);

psi_ds1(w)=psi_ds1_int+dt*(vds1(w-1)-Rs1*ids1(w-1)...+omegae1(w-1)*psi_qs1(w-1));

psi_ds1_int=psi_ds1(w);

psi_qr1(w)=psi_qr1_int+dt*(vqr1-Rr1*iqr1(w-1)...-(omegae1(w-1)-2*thetadot_1(w-1))*psi_dr1(w-1));

psi_qr1_int=psi_qr1(w);

psi_dr1(w)=psi_dr1_int+dt*(vdr1-Rr1*idr1(w-1)...+(omegae1(w-1)-2*thetadot_1(w-1))*psi_qr1(w-1));

psi_dr1_int=psi_dr1(w);

%calculate dq currentspsi1=[psi_qs1(w);psi_ds1(w);psi_qr1(w);psi_dr1(w)];I1=inv(Lmat1)*psi1;iqs1(w)=I1(1);ids1(w)=I1(2);iqr1(w)=I1(3);idr1(w)=I1(4);

%calculate developed torquetorque_1(w)=(3/2)*(2)*(Lm1)...

*(iqs1(w)*idr1(w)-ids1(w)*iqr1(w));

%***drive 5%determine rotating field frequencyomegae5_integral=dt*((K_DR5/tau_DR5)*tension_ctrl_out(w-1))...

-dt*((1/tau_DR5)*omegae5(w-1));

176

omegae5(w)=omegae5_integral+intsum_omegae5;intsum_omegae5=omegae5(w);

%develop pwm voltage signalsdrive5_thetas(w)=drive5_thetas_int+dt*omegae5(w);drive5_thetas_int=drive5_thetas(w);

if abs(omegae5(w))<(6*2*pi)llvolts=VpHzegain5*exp(VpHzexp5*(omegae5(w)/(2*pi)));voltmag_5=llvolts/sqrt(3);

elseif abs(omegae5(w))>=(6*2*pi)llvolts=VpHzgain5*(omegae5(w)/(2*pi))+VpHzoffset5;voltmag_5=llvolts/sqrt(3);

end

%COAST COMMANDif time(w)>=t_sagstart+0.025 & time(w)<=t_sagend+0.002

coastcommand5(w)=1;end

%DRIVE DROPOUT SEQUENCE AND DC BUS DECAY%SECOND CONDITIONAL INCLUDED IN RECOVERYif time(w)>=t_sagstart & time(w)<=t_sagend

if dcbus5(w-1)>=drivetrip5dcbus5(w)=dcbus5(w-1)-dcdecayA*dt;

endif dcbus5(w-1)<drivetrip5 | coastcommand5(w)==1

dcbus5(w)=dcbus5(w-1)-dcdecayB*dt;driveout5(w)=0;voltmag_5=0;psi_qs5(w-1)=0;psi_qr5(w-1)=0;psi_ds5(w-1)=0;psi_dr5(w-1)=0;

endend

%permanent trip after loss of driveif trip5==0 & dcbus5(w)<drivetrip5

trip5=1;end

%if driveout5(w-1)==0; %maintains trip for no mitigationif trip5==1;

driveout5(w)=0;voltmag_5=0;psi_qs5(w-1)=0;psi_qr5(w-1)=0;psi_ds5(w-1)=0;psi_dr5(w-1)=0;

end

vas5=voltmag_5*sqrt(2)*cos(drive5_thetas(w));vbs5=voltmag_5*sqrt(2)*cos(drive5_thetas(w)-2*pi/3);

177

vcs5=voltmag_5*sqrt(2)*cos(drive5_thetas(w)+2*pi/3);


vqs5(w)=vqss5*cos(drive5_thetas(w))-vdss5*sin(drive5_thetas(w));vds5(w)=vqss5*sin(drive5_thetas(w))+vdss5*cos(drive5_thetas(w));

%calculate dq fluxespsi_qs5(w)=psi_qs5_int+dt*(vqs5(w-1)-Rs5*iqs5(w-1)...

-omegae5(w-1)*psi_ds5(w-1));psi_qs5_int=psi_qs5(w);

psi_ds5(w)=psi_ds5_int+dt*(vds5(w-1)-Rs5*ids5(w-1)...+omegae5(w-1)*psi_qs5(w-1));

psi_ds5_int=psi_ds5(w);

psi_qr5(w)=psi_qr5_int+dt*(vqr5-Rr5*iqr5(w-1)...-(omegae5(w-1)-2*thetadot_5(w-1))*psi_dr5(w-1));

psi_qr5_int=psi_qr5(w);

psi_dr5(w)=psi_dr5_int+dt*(vdr5-Rr5*idr5(w-1)...+(omegae5(w-1)-2*thetadot_5(w-1))*psi_qr5(w-1));

psi_dr5_int=psi_dr5(w);

%calculate dq currentspsi5=[psi_qs5(w);psi_ds5(w);psi_qr5(w);psi_dr5(w)];I5=inv(Lmat5)*psi5;iqs5(w)=I5(1);ids5(w)=I5(2);iqr5(w)=I5(3);idr5(w)=I5(4);

%calculate developed torquetorque_5(w)=(3/2)*(2)*(Lm5)...

*(iqs5(w)*idr5(w)-ids5(w)*iqr5(w));%----------------------------------------------------------------%

%----------------------------------------------------------------%%mechanical system

%assemble input matrix from motor developed torquesu1=torque_1(w);u2=torque_5(w);

%calculate output matrix y=Cx+Du%x is initialized for w=1, recalculated for next iteration latery=C*x+D*[u1;u2];

%determine linespeed and tension variables from output matrixlinespeed(w)=y(1);thetadot_1(w)=y(1)/r1;thetadot_5(w)=y(2)/r5;pten_actual(w)=y(3);

178

sten_actual(w)=y(4);

%slackif pten_actual(w) < 0

pten_actual(w) = 0;end

if sten_actual(w) < 0sten_actual(w) = 0;

end

%calculate x for next iterationx=x_intsum+dt*(A*x+B*[u1;u2]);x_intsum=x;

%progress messagesif rem(w,10000)==0

wclock

end%----------------------------------------------------------------%

%loop endend%--------------------------------------------------------------------%%END PROGRAM SAGSIM.M

179

Appendix E

Instrumentation Voltage Sag Response Data

This appendix contains the raw data obtained from sag testing the textile tension

controller analog instrumentation.. All values are given in seconds. Voltage sag

durations for ridethrough time tests were fixed at 450ms. Magnitudes were varied

from 0% to 80% in 20% increments. When the sensor rode through the event without

an observable disturbance, ‘RT’ is indicated. Maximum recovery times found during

the test session were also recorded. Sag durations were varied up to 450ms to ensure

the maximum recovery time was obtained.

Table E.1 Ridethrough times obtained from sag testing of tension sensor.

PHYS. INPUT V=0% V=20% V=40% V=60% V=80%5N 0.085 0.074 0.074 0.074 RT14N 0.274 0.274 RT RT RT23N 0.150 0.147 0.136 0.124 RT32N 0.078 0.075 0.072 0.076 RT41N 0.060 0.060 0.064 0.064 RT50N 0.060 0.060 0.060 0.060 RT

Table E.2 Ridethrough times obtained from sag testing of payoff tachometer.

PHYS. INPUT V=0% V=20% V=40% V=60% V=80%500 mm/s 0.096 0.096 0.096 RT RT1000 mm/s 0.090 0.090 0.090 RT RT1500 mm/s 0.086 0.086 0.086 RT RT2000 mm/s 0.080 0.080 0.080 RT RT2500 mm/s 0.076 0.076 0.076 RT RT

Table E.3 Ridethrough times obtained from sag testing of takeup tachometer.

PHYS. INPUT V=0% V=20% V=40% V=60% V=80%500 mm/s 0.094 0.094 0.094 RT RT1000 mm/s 0.086 0.086 0.086 RT RT1500 mm/s 0.080 0.080 0.080 RT RT2000 mm/s 0.080 0.080 0.080 RT RT2500 mm/s 0.076 0.076 0.076 RT RT

180

Table E.4 Maximum recovery times obtained from instrument sag tests.

INSTRUMENT MAX RECOVERY TIMEtension cell 0.088

payoff tachometer 0.246takeup tachometer 0.246

181

Appendix F

AC Motor Drive Voltage Sag Response Data

This appendix contains AC motor drive ridethrough results for sags of varying

magnitudes and phasing combinations, at a duration of 450ms. Sag magnitudes

are expressed as percentages in phase order A-B-C. Combinations marked with a ‘*’

indicate a dropout, where the drive halted its output during the sag.

182

Table F.1 AC motor drive sag response data.

(0,0,0)* (20,0,0)* (40,0,0)*(0,0,20)* (20,0,20)* (40,0,20)*(0,0,40)* (20,0,40)* (40,0,40)*(0,0,60)* (20,0,60)* (40,0,60)*(0,0,80)* (20,0,80)* (40,0,80)*(0,0,100)* (20,0,100)* (40,0,100)*(0,20,0)* (20,20,0)* (40,20,0)*(0,20,20)* (20,20,20)* (40,20,20)*(0,20,40)* (20,20,40)* (40,20,40)*(0,20,60)* (20,20,60)* (40,20,60)*(0,20,80)* (20,20,80)* (40,20,80)*(0,20,100)* (20,20,100)* (40,20,100)*(0,40,0)* (20,40,0)* (40,40,0)*(0,40,20)* (20,40,20)* (40,40,20)*(0,40,40)* (20,40,40)* (40,40,40)*(0,40,60)* (20,40,60)* (40,40,60)*(0,40,80)* (20,40,80)* (40,40,80)*(0,40,100)* (20,40,100)* (40,40,100)*(0,60,0)* (20,60,0)* (40,60,0)*(0,60,20)* (20,60,20)* (40,60,20)*(0,60,40)* (20,60,40)* (40,60,40)*(0,60,60)* (20,60,60)* (40,60,60)*(0,60,80)* (20,60,80)* (40,60,80)*(0,60,100) (20,60,100) (40,60,100)(0,80,0)* (20,80,0)* (40,80,0)*(0,80,20)* (20,80,20)* (40,80,20)*(0,80,40)* (20,80,40)* (40,80,40)*(0,80,60)* (20,80,60)* (40,80,60)*(0,80,80) (20,80,80) (40,80,80)(0,80,100) (20,80,100) (40,80,100)(0,100,0)* (20,100,0)* (40,100,0)*(0,100,20)* (20,100,20)* (40,100,20)*(0,100,40)* (20,100,40)* (40,100,40)*(0,100,60) (20,100,60) (40,100,60)(0,100,80) (20,100,80) (40,100,80)(0,100,100) (20,100,100) (40,100,100)

183

Table F.2 AC motor drive sag response data (continued).

(60,0,0)* (80,0,0)* (100,0,0)*(60,0,20)* (80,0,20)* (100,0,20)*(60,0,40)* (80,0,40)* (100,0,40)*(60,0,60)* (80,0,60)* (100,0,60)(60,0,80)* (80,0,80) (100,0,80)(60,0,100) (80,0,100) (100,0,100)(60,20,0)* (80,20,0)* (100,20,0)*(60,20,20)* (80,20,20)* (100,20,20)*(60,20,40)* (80,20,40)* (100,20,40)*(60,20,60)* (80,20,60)* (100,20,60)(60,20,80)* (80,20,80) (100,20,80)(60,20,100) (80,20,100) (100,20,100)(60,40,0)* (80,40,0)* (100,40,0)*(60,40,20)* (80,40,20)* (100,40,20)*(60,40,40)* (80,40,40)* (100,40,40)*(60,40,60)* (80,40,60)* (100,40,60)(60,40,80)* (80,40,80) (100,40,80)(60,40,100) (80,40,100) (100,40,100)(60,60,0)* (80,60,0)* (100,60,0)(60,60,20)* (80,60,20)* (100,60,20)(60,60,40)* (80,60,40)* (100,60,40)(60,60,60)* (80,60,60)* (100,60,60)(60,60,80)* (80,60,80) (100,60,80)(60,60,100) (80,60,100) (100,60,100)(60,80,0)* (80,80,0) (100,80,0)(60,80,20)* (80,80,20) (100,80,20)(60,80,40)* (80,80,40) (100,80,40)(60,80,60)* (80,80,60) (100,80,60)(60,80,80) (80,80,80) (100,80,80)(60,80,100) (80,80,100) (100,80,100)(60,100,0) (80,100,0) (100,100,0)(60,100,20) (80,100,20) (100,100,20)(60,100,40) (80,100,40) (100,100,40)(60,100,60) (80,100,60) (100,100,60)(60,100,80) (80,100,80) (100,100,80)(60,100,100) (80,100,100) (100,100,100)

184

Appendix G

LabView Code For Data Read and Profibus Functions

This appendix contains LabView code for the data file read and Profibus interface

functions used in the Ridethrough PC.

Figure G.1 Labview subroutine to close Profibus interface.

185

Figure G.2 Labview subroutine to read in data file for ridethrough PC (part 1 of 2).

Figure G.3 Labview subroutine to read in data file for ridethrough PC (part 2 of 2).

186

Figure G.4 Labview subroutine for Profibus open interface (part 1 of 2).

187

Figure G.5 Labview subroutine for Profibus open interface (part 2 of 2).

188

Appendix H

PLC Ladder Code

This appendix contains the ladder code used in the textile tension control process

PLC and the added subroutine used to interface with the Ridethrough PC.

189

Appendix I

Additional Ridethrough PC Software Flowcharts

This appendix contains additional flowcharts detailing the software structure for

the Ridethrough PC program. They are provided as an addendum to the flowcharts

given in Chapter 3, “Software Design”.

START MAIN

DEFINE INSTRUMENTRIDETHROUGH AND

RECOVERY TIMEARRAY VALUES

INITIALIZE EVENTHANDLING ARRAY

INITIALIZE:3 MIN RMS=1,3 MIN dRMS=0,

START ITERATION=0,END ITERATION=0,EVENT LENGTH=0,

8 REPLACEMENT SIGNALS=0,VOLTAGE SIGNAL ARRAYS,

IS_SAG=FALSE,WAS_SAG=FALSE

USER SELECTED READFROM DATA FILE?

YES

NO

DATA FILE INPUTAND

RESTRUCTURINGSUBROUTINE

INITIALIZE PROFIBUSINTERFACE

EVENT COUNTER=-1

EVENT COUNTER=SIZE(STORED ARRAY)

DEFINE:DATA ACQUISITION

CARD ID,INPUT CHANNELS,

INPUT BUFFER SIZE,SCAN RATE,

NUMBER OF SCANS PERPROGRAM CYCLE=5

MAIN WHILELOOP

USER SELECTED WRITETO DATA FILE?

YES

NO

END MAIN

WRITE / APPENDDATA FILE

SUBROUTINE

CLOSE PROFIBUSINTERFACE

Figure I.1 Flowchart for main program of ridethrough PC.

208

START OPENINTERFACE

END OPENINTERFACE

LOAD DRIVER

OPEN CARDCOMMUNICATIONS

RESET INTERFACECARD

CONFIGURE CARDAS SLAVE DEVICE

CONFIGURENETW ORKINTERFACE

GO ONLINE WITHNETW ORK

Figure I.2 Flowchart for Profibus open interface routine.

START CLOSEINTERFACE

END CLOSEINTERFACE

GO OFFLINE WITHNETWORK

UNLOAD DRIVERFROM MEMORY

Figure I.3 Flowchart for Profibus close interface routine.

209

START DATAFILE READ

END DATAREAD

OPEN TEXTDATA FILE

INTITIALIZE 6X6X6EVENT ARRAY

END OF FILEREACHED?

YES

NO

READ DATA FORSINGLE

RECORDEDEVENT

CALCULATE ARRAYINDICES FOR NEXT

EVENT IN RAW DATAARRAY

END OF RAW DATA REACHED?

YES

NO

PLACE EVENT DATAIN APPROPRIATE

6X6X6 ARRAYLOCATION

Figure I.4 Flowchart for data file read routine.

210

START DATAFILE WRITE

END OF RAW DATA REACHED?

YES

NO

WRITE DATA FORSINGLE EVENT IN

TEXT FILE

END DATA FILEWRITE

Figure I.5 Flowchart for data file write routine.

211

USER TERMINATEDPROGRAM?

YES

NO

START MAINWHILE LOOP

TERMINATELOOP

END MAIN WHILELOOP

VOLTAGEMEASURE ANDSAG DETECT

DATA BUS READAND WRITE

EVENTANALYSIS

RESTART /RESET

SELECTIVESIGNAL

SUBSTITUTION

CONTROLSIGNAL

SUBSTITUTION

HARDWARECOAST

SOFTWARECOAST

MEASUREMENTRESPONSE ID

HISTORYRESPONSE ID

Figure I.6 Flowchart for main loop of ridethrough PC software.

212

START BUSREAD / WRITE

INSERT NEWPROFIBUS DATA

INTO ARRAYS

DISCARD OLDESTBUS DATA POINTFROM ARRAYS

READ 16 BYTEPROFIBUS

DATA - STATUSAND SIGNALS

WRITE 16 BYTEPROFIBUS DATA -

CONTROL ANDREPLACEMENT

SIGNALS

END BUS READ /WRITE

Figure I.7 Flowchart for Profibus read and write routine.

213

Appendix J

Main LabView Code For Ridethrough PC

This appendix contains the main LabView code used in the Ridethrough PC.

Structure windows that appear to be disconnected are secondary cases of the con-

nected windows immediately adjacent to them.

Figure J.1 Main Labview code - variable initialization before main loop.

214

Figure J.2 Main LabView code - data read, array initialization, and hardware setupbefore main loop.

215

Figure J.3 Main LabView code - Profibus I/O and data monitoring.

216

Figure J.4 Main LabView code - event data capture.

217

Figure J.5 Main LabView code - prefault average calculation.

218

Figure J.6 Main LabView code - data acquisition and voltage data handling.

219

Figure J.7 Main LabView code - voltage sag detection, minimum RMS voltage, andminimum RMS voltage derivative determination.

220

Figure J.8 Main LabView code - event analysis instructions.

221

Figure J.9 Main LabView code - peak voltage detection instructions.

Figure J.10 Main LabView code - expected process response identification usingrecorded process history.

222

Figure J.11 Main LabView code - expected process response identification using for-mal region boundary definitions.

Figure J.12 Main LabView code - software coast agorithm instruction set.

223

Figure J.13 Main LabView code - hardware coast algorithm instruction set.

Figure J.14 Main LabView code - restarting instruction set.

224

Figure J.15 Main LabView code - control signal substitution and status clearinginstruction set.

Figure J.16 Main LabView code - generation of ridethrough time index for selectivesignal substitution algorithm.

225

Figure J.17 Main LabView code - selective signal substitution algorithm replace bitset and reset instructions.

226

Figure J.18 Main LabView code - data file write and Profibus interface close functioncall after termination of main loop.

227

BIBLIOGRAPHY

[1] “Solving the Fast Clock Problem”, Power Quality Testing Network Solutions,Knoxville, TN, The EPRI Power Electronics Applications Center, No.3,September 1994.

[2] “Electronic Digital Clock Performance During Steady-State and Dynamic Power Dis-turbances”, Power Quality Testing Network Briefs, Knoxville, TN, TheEPRI Power Electronics Applications Center, No.17, March 1994.

[3] “Undervoltage Ride-Through Performance of Off-the-Shelf Personal Computers”,Power Quality Testing Network Briefs, Knoxville, TN, The EPRI PowerElectronics Applications Center, No.7, December 1992.

[4] “Low-Voltage Ride-Through Performance of a Personal Computer Power Supply”,Power Quality Testing Network Briefs, Knoxville, TN, The EPRI PowerElectronics Applications Center, No.11, March 1993.

[5] “IEEE Recommended Practice for Monitoring Electric Power Quality”, IEEE Stan-dard 1159-1995, New York, The Institute of Electrical and Electronics En-gineers, Inc., 1995.

[6] “Power Quality Considerations for Adjustable Speed Drives”, EPRI, 1991.

[7] “IEEE Recommended Practices and Requirements for Harmonic Control in ElectricalPower Systems”, IEEE Standard 519-1992, New York, The Institute ofElectrical and Electronics Engineers, Inc., 1992.

[8] “Article 517 - Health Care Facilities”, The National Electrical Code, 2008 Edition,Quincy, MA, National Fire Prevention Association, Inc., 2008, pp.425-442.

[9] John P. Wagner. “The Cost of Power Quality in the ITE Industry”, Proceedings ofthe Second International Conference on Power Quality, PQA, ’92: End UseApplications and Perspectives, Atlanta, GA, EPRI, 1992, pp.A-3:1-A-3:5.

[10] Ambra Sannino, Michelle Ghans Miller, Math H.J. Bollen. “Overview of VoltageSag Mitigation”, Proceedings of the IEEE 2000 Power Engineering SocietyWinter Meeting, Vol. 4., January 2000, pp.2872-2878.

[11] Michael J. Sullivan, Terry Vardell, Mark Johnson. “Power Interruption Costs to In-dustrial and Commercial Consumers of Electricity” IEEE Transactions onIndustry Applications, Vol.33, No.6, November/December 1997, pp.1448-1458.

[12] Marek Samotyj. “Solutions to Voltage Sag Problems”, EPRI Journal, Vol.20, No.4,July/August 1995, pp.42-45.

[13] Mark F. McGranaghan, David R. Mueller, Marek J. Samotyj. “Voltage Sags in Indus-trial Systems” IEEE Transactions on Industry Applications, Vol.29, No.2,March/April 1993, pp.397-402.

228

[14] Hamish Laird, Simon Round, Richard Duke, Alister Gardiner. “Power Quality Obser-vations at a Light Industrial Site”, Proceedings of the IEEE Eighth Inter-national Conference on Harmonics and Quality of Power, Athens, Greece,October 1998, pp.88-93.

[15] Van Wagner, Thomas Grebe, Robert Kretschmann, Lawrence Morgan, Al Price.“Power System Compatibility with Industrial Process Equipment”, IEEEIndustry Applications Magazine, Vol.2, No.1, January/February 1996.

[16] J.V. Milanovic, R. Gnativ, K.W.M. Chow. “The Influence of Loading Conditionsand Network Topology on Voltage Sags”, Proceedings of the IEEE NinthInternational Conference on Harmonics and Quality of Power, Orlando,FL, October 2000, pp.757-762.

[17] P.R. Chaney. The Characterization of Voltage Sags on an Electric Power System,Clemson, SC, Clemson University M.S. Thesis, 1998.

[18] Math H.J. Bollen. “The Influence of Motor Reacceleration on Voltage Sags”, IEEETransactions on Industry Applications, Vol.31, No.4, July/August 1995,pp.667-674.

[19] Larry Conrad, Kevin Little, Cliff Grigg. “Predicting and Preventing Problems Asso-ciated with Remote Fault-Clearing Voltage Dips”, IEEE Transactions onIndustry Applications, Vol.27, No.1, January/February 1991, pp.167-172.

[20] Waveform Characterization of Voltage Sags: Definition and Algorithm Development,Palo Alto, CA, EPRI, 1999.

[21] “Power Quality Considerations for the Textile Industry”, EPRI & Duke Power Com-pany, 1995.

[22] Linda S. Tillis, Charles W. Williams, Jr.. “Voltage Sag Investigation and Mitigationin an Electronics Manufacturing Plant”, IEEE Southcon ’94 ConferenceRecord, Orlando, FL, March 1994, pp.378-381.

[23] C.A. Warren, T.A. Short, J.J. Burke, H. Morosini, C.W. Burns, J. Storms. “PowerQuality at Champion Paper - The Myth and the Reality”, IEEE Transac-tions on Power Delivery, Vol.14, No.2, April 1999, pp.636-641.

[24] Jeff Lamoree, Dave Mueller, Paul Vinett, William Jones, Marek Samotyj. “VoltageSag Analysis Case Studies” IEEE Transactions on Industry Applications,Vol.30, No.4, July/August 1994, pp.1083-1089.

[25] “Low Voltage Ride-Through Performance of 5-HP Adjustable-Speed Drives”, PowerQuality Testing Network Briefs, Knoxville, TN, The EPRI Power Electron-ics Applications Center, No.9, January 1993.

[26] “Low Voltage Ride-Through Performance of AC Contactor Motor Starters”, PowerQuality Testing Network Briefs, Knoxville, TN, The EPRI Power Electron-ics Applications Center, No.10, February 1993.

[27] E.R. Collins, M.A. Bridgwood. “The Impact of Power Disturbances on AC-Coil Con-tactors,” Proceedings of the 1997 IEEE Textile, Fiber, and Film IndustryTechnical Conference, Greenville, SC, 1997, pp.2-7

229

[28] Programmable Logic Controllers used in Industrial Power Systems, PLC I Final Re-port, Knoxville, TN, EPRI Power Electronics Applications Center, August1996.

[29] “Ride-Through Performance of Programmable Logic Controllers”, Power QualityTesting Network Briefs, Knoxville, TN, The EPRI Power Electronics Ap-plications Center, No.39, November 1996.

[30] “IEEE Recommended Practice for Evaluating Electric Power System Compatibilitywith Electronic Process Equipment”, IEEE Standard 1346-1998, New York,The Institute of Electrical and Electronics Engineers, Inc., 1998.

[31] Larry E. Conrad, Math H.J. Bollen. “Voltage Sag Coordination for Reliable PlantOperation” IEEE Transactions on Industry Applications, Vol.33, No.6, No-vember/December 1997, pp.1459-1464.

[32] ITI(CBEMA) Curve, Published by Information Technology Industry Coun-cil (ITI), 1250 Eye Street NW, Suite 200, Washington, DC 20005,http://www.itic.org.

[33] S. Middlekauff. Using a Sag-Generator to Identify Power Quality Problems in Indus-try, Clemson, SC, Clemson University M.S. Thesis, 1996.

[34] E.R. Collins, Jr., R.L. Morgan. “A Three-Phase Sag Generator for Testing IndustrialEquipment”, IEEE Transactions on Power Delivery, Vol.11, No.1, January1996, pp.526-532.

[35] R. Caldon, M. Fauri, L. Fellin. “Voltage Sag Effects on Continuous IndustrialProcesses: Desensitizing Study for Textile Manufacture”, Proceedings ofthe Second International Conference on Power Quality, PQA, ’92: EndUse Applications and Perspectives, Atlanta, GA, EPRI, 1992, pp.D-13:1-D-13:7.

[36] M.A. Bridgwood, J. Spangler, P. Miroftsalis, E.R. Collins. “Characterizing the Behav-ior of a Multi-Loop Model Industrial Process Subjected to Complex PowerSupply Voltage Sags”, Proceedings of the Eleventh International PowerQuality ’98 Symposium, Santa Clara, CA, November 1998, pp.414-425.

[37] M.A. Bridgwood, O.C. Parks, E.R. Collins. “Using Macromodels to Evaluate aProcess Control System’s Response to Voltage Sags”, Proceedings of theEleventh International Conference on Harmonics and Quality of Power,Lake Placid, NY, September 2004, pp.717-720.

[38] Carlos Alvarez, Javier Alamar, Alex Domijan Jr., Alejandro Montenego, ZhidongSong. “An Investigation Toward New Technologies and Issues in PowerQuality”, Proceedings of the IEEE Ninth International Conference on Har-monics and Quality of Power, Orlando, FL, October 2000, pp.444-455.

[39] Kay Tuttle. “How to Position Power Quality Within a Utility”, Proceedings of theIEEE Ninth International Conference on Harmonics and Quality of Power,Orlando, FL, October 2000, pp.438-443.

[40] Thomas M. Gruzs. “Power Quality - A Shared Responsibility”, Proceedings of theFirst International Power Quality Conference - Power Quality ’89 Sympo-sium, Long Beach, CA, October 1989, pp.32-41.

230

[41] Douglas S. Dorr, M. Brent Hughes, Thomas M. Gruzs, Robert E. Jurewicz, John L.McClaine. “Interpreting Recent Power Quality Surveys to Define the Elec-trical Environment” IEEE Transactions on Industry Applications, Vol.33,No.6, November/December 1997, pp.1480-1487.

[42] Melisa Johns, Larry Morgan. “Voltage Sag Mitigation Through Ridethrough Coor-dination”, Proceedings of 1994 IEEE/IAS Annual Textile, Fiber and FilmIndustry Technical Conference, Greenville, SC, May 1994, pp.1-5.

[43] Robert C. Degeneff, Russ Barss, Danial Carnovale, Steven Raedy. “Reducing the Ef-fect of Sags and Momentary Interruptions: A Total Owning Cost Prospec-tive”, Proceedings of the IEEE Ninth International Conference on Harmon-ics and Quality of Power, Orlando, FL, October 2000, pp.397-409.

[44] Math H.J. Bollen, Thavatchai Tayjasanant, Gulali Yalcinkaya. “Assessment of theNumber of Voltage Sags Experienced by a Large Industrial Customer” IEEETransactions on Industry Applications, Vol.33, No.6, November/December1997, pp.1465-1471.

[45] Milind M. Bhanoo. “Static Transfer Switch: Advances in High Speed Solid-StateTransfer Switches for Critical Power Quality and Reliability Applications”,Proceedings of the 1998 IEEE Annual Textile, Fiber, and Film IndustryTechnical Conference, May 1998, pp.5/1-5/8.

[46] Roger C. Dugan, Mark F. McGranaghan, H. Wayne Beaty. Electrical Power SystemsQuality, Second Edition, New York, McGraw-Hill, 1996.

[47] H.J. Kim. A New Approach to Correct Power System Voltage Disturbances Using aSix Pulse Inverter, Clemson, SC, Clemson University M.S. Thesis, 2000.

[48] P. Wang, N. Jenkins, M.H.J. Bollen. “Experimental Investigation of Voltage SagMitigation By an Advanced Static Var Compensator” IEEE Transactionson Power Delivery, Vol.13, No.4, October 1998, pp.1461-1467.

[49] “Ferro-Resonant Transformer Output Performance Under Varying Supply Condition-s”, Power Quality Testing Network Briefs, Knoxville, TN, The EPRI PowerElectronics Applications Center, No.13, May 1993.

[50] “Ride-Through Performance of a Web Process Enhanced by a Constant-VoltageTransformer”, Power Quality Testing Network Briefs, Knoxville, TN, TheEPRI Power Electronics Applications Center, No.40, November 1996.

[51] S.N. Tunaboylu. Electromechanical Performance of Induction Motor/Drive SystemsSubjected to Voltage Sags, Clemson, SC, Clemson University Ph.D. Disser-tation, 2002.

[52] “Low Voltage Ride-Through Performance of a Modified Personal Computer PowerSupply”, Power Quality Testing Network Briefs, Knoxville, TN, The EPRIPower Electronics Applications Center, No.12, April 1993.

[53] M.A. Bridgwood, P. Miroftsalis, E.R. Collins, J. Spangler. “Sag Injection/DetectionSystems for PLC-Based Industrial Process Control Applications”, Proceed-ings of the Eleventh International Power Quality ’98 Symposium, SantaClara, CA, November 1998, pp.511-519.

231

[54] N.T. Youell. Development of a Real-Time, Three-Phase, Cost-Effective Voltage Dis-turbance Detector, Clemson, SC, Clemson University M.S. Thesis, 2001.

[55] Oscar C. Montero-Hernandez, Prasad N. Enjeti. ‘A Fast Detection Algorithm Suit-able for Mitigation of Numerous Power Quality Disturbances’, ConferenceRecord of the 36th IEEE Industry Applications Society (IAS) Annual Meet-ing, Vol.4, October 2001, pp.2661-2666.

[56] J. Ackermann. ‘Robustness Against Sensor Failures’, Automatica, Vol.20, No.2, March1984, pp.211-215.

[57] Koicha Suyama. ‘Fault Detection of Redundant Sensors used in Reliable Sampled-Data Control Systems’ Proceedings of the 37th IEEE Conference on Deci-sion and Control, Tampa, FL, December 1998, Vol.1, pp.1161-1164.

[58] G. Hearns, M.J. Grimble. ‘Fault Tolerant Strip Tension Control’, Proceedings of theAmerican Control Conference, Philadelphia, PA, June 1998, pp.2992-2996.

[59] W. Wolfermann. ‘Sensorless Tension Control of Webs’, Proceedings of the FourthInternational Conference on Web Handling, June 1997, Web Handling Re-search Center, Oklahoma State University, Stillwater, OK, pp.318-340.

[60] Seung-Ho Song, Seung-Ki Sul. ‘A New Tension Controller for Continuous StripProcessing Line’, IEEE Transactions on Industry Applications, Vol.36,No.2, March/April 2000, pp.633-639.

[61] Norbert A. Ebler, Ragnar Arnason, Gerd Michaelis, Noel D’Sa. ‘Tension Control:Dancer Rolls or Load Cells’, IEEE Transactions on Industry Applications,Vol.29, No.4, July/August 1993, pp.727-739.

[62] J. Timothy Shea. ‘U.S. Device/Sensor Bus Race: Three Contestants Outpace theField’, Instrumentation and Control Systems, Vol.71, No.6, June 1998,pp.61-65.

[63] B.D. Van Blerk, G. Diana. ‘Development of a Generic Tension Control System’,AFRICON 1996, IEEE 4th AFRICON, September 1996, Vol.2, pp.864-868.

[64] Priyadarshee D. Mathur, William C. Messner. ‘Controller Development for a Pro-totype High-Speed Low-Tension Tape Transport’, IEEE Transactions onControl Systems Technology, Vol.6, No.4, July 1998, pp.534-542.

[65] Bimal K. Bose. Modern Power Electronics and AC Drives, Upper Saddle River, NJ,Prentice-Hall, 2001.

[66] Alex McEachern. “Designing Electronic Devices to Survive Power Quality Events”,IEEE Industry Applications Magazine, Vol. 6, No. 6, November/December2000, p.66-9.

[67] Duke Energy Process Sensors Test Study, Prepared by EPRI PEAC Corporation,2000.

[68] J.T. Spangler. Forming the Mathematical Model of a Multiple Loop Feedback ControlsSystem for Improving the Ride-Through Capabilities of a Complex TextileProcess, Clemson, SC, Clemson University M.S. Thesis, 1999.

232

[69] O.C. Parks. The Development of a Serial Data Bus Driven Textile Process for Usein Power Quality Studies, Clemson, SC, Clemson University M.S. Thesis,2000.

[70] Ferdinand P. Beer, E. Russell Johnston, Jr.. Vector Mechanics for Engineers: Dy-namics, New York, McGraw-Hill, 1984.

[71] William L. Brogan. Modern Control Theory, Third Edition, Upper Saddle River, NJ,Prentice-Hall, 1991.

[72] “Test Designation ASTM D5035-95”, Annual Book of ASTM Standards, Vol 7.01,West Conoshocken, PA, American Society for Testing and Materials, 1999.

[73] Stephen J. Chapman. Electric Machinery Fundamentals, Second Edition, New York,McGraw-Hill, 1991.

[74] David G. Hart, William Peterson, David Uy, Jochen Schneider, Damir Novosel, RayWright. ‘Tapping Protective Relays for Power Quality Information’ IEEEComputer Applications in Power, Vol.13, Issue 1, January 2000, pp.45-9

[75] Mesut E. Baran, Jinsang Kim, David G. Hart, David Lubkeman, Glenn C. Lampley,William F. Newell. ‘Voltage Variation Analysis for Site-Level PQ Assess-ment’, IEEE Transactions on Power Delivery, Vol.19, No.4, October 2004,pp.1956-1961.

[76] Roy Moxley. ‘Synchrophasors in the Real World’, Pullman, WA, Schweitzer Engineer-ing Laboratories Inc., http://www.selinc.com, 2005.

[77] Mark Adamiak, William Premerlani, Bogdan Kasztenny. ‘Synchrophasors: Defini-tion, Measurement, and Application’, GE Corporation (GE Multilin/GECorporate Research), http://www.gedigitalenergy.com, 2009.

233

Software Algorithms to Coordinate and Improve Voltage Sag ...

Documents

Transcript of Software Algorithms to Coordinate and Improve Voltage Sag ...